Studying Cosmic Evolution with the XMM-Newton Distant Cluster Project: X-Ray Luminous Galaxy Clusters at Z>~ 1 and Their Galaxy Populations

Home , Abell 133, Abell 1835, Cosmic Evolution Survey, EX Hydrae, HD 16004, NGC 1132

arXiv:0806.0861v1 [astro-ph] 4 Jun 2008 M-etnDsatCutrProject: Cluster Distant XMM-Newton e Ludwig-Maximilians-Universität Münchender tdigCsi vlto ihthe with Evolution Cosmic Studying -a uiosGlx lsesa z at Clusters Galaxy Luminous X-ray israindrFakultät für der Physik Dissertation n hi aayPopulations Galaxy their and eeFassbender Rene ¨ nhn koe 2007 München, Oktober u Diez/Lahn aus oglg von vorgelegt ∼ > 1 1. Gutachter: Prof. Dr. Hans Böhringer 2. Gutachter: Prof. Dr. Andreas Burkert

Tag der m¨undlichen Pr¨ufung: 29. November 2007 Dedicated to the memory of Peter Schuecker (1959–2006)

Beobachtung Kosmischer Entwicklung mit dem XMM-Newton Distant Cluster Project: Röntgenhelle Galaxienhaufen bei z>1 und ihre Galaxienpopulationen ∼ Zusammenfassung Röntgenhelle Galaxienhaufen bei hoher Rotverschiebung (z >1) ermöglichen einzigartige, aber herausfordernde, Untersuchungen der kosmischen Entwicklung∼ dieser größten Objekte im Uni- versum, der baryonischen Materie in Form des heißen Haufengases (Intracluster Medium, ICM), ihrer Galaxienpopulationen und des Einflusses der mysteriösen Dunklen Energie. Das Hauptziel der vorliegenden Doktorarbeit ist das Erstellen der Datengrundlage für das XMM- Newton Distant Cluster Project (XDCP). Diese neue Himmelsdurchmusterung im Röntgenlicht konzentriert sich auf die systematische Suche und Identifizierung von Galaxienhaufen bei den höchsten Rotverschiebungen von z > 1 mittels einer gezielten Vorauswahl von ausgedehnten Röntgenquellen, Nachbeobachtungen∼ in zwei photometrischen Bändern zum Nachweis und zur Entfernungsschätzung der Objekte sowie der abschließenden spektroskopischen Bestätigung. Als ersten Schritt habe ich insgesamt 80 Quadratgrad (469 Felder) langbelichteter Röntgenbeobachtungen aus dem XMM-Newton Datenarchiv mit einem neu entwickelten au- tomatisierten Datenreduktionssystem analysiert. Damit konnte ich annähernd 1 000 aus- gedehnte Quellen als Galaxienhaufen-Kandidaten nachweisen, von denen sich 75% durch vorhandene optische Daten als Haufen und Gruppen bei z < 0.6 identifizieren ließen. Die verbleibenden ca. 250 Kandidaten mit typischen Röntgenfl¨∼ussen von 10−14 ergs−1cm−2 ≃ (0.5–2.0 keV) erfordern weitere Schritte, um als entfernte Galaxienhaufen bestätigt werden zu können. Dafür entwickelte ich eine neue Nahinfrarot-Beobachtungsstrategie, mit der sich die Iden- tifizierung der ausgedehnten Röntgenquellen mittels Z- und H-Band Aufnahmen effizient durchführen lässt und die zur Rotverschiebungsabschätzung anhand der Z–H Farbe der ‘Red- Sequence’ der Frühtypgalaxien dient. Um das Potential dieser Methode voll auszuschöpfen, konzipierte ich ein neues Nahinfrarot-Datenverarbeitungsprogramm, mit dem die Nachbeobach- tungen zweier Kampagnen mit dem 3.5 m Teleskop des Calar Alto Observatoriums für 25% der 250 entfernten Galaxienhaufen-Kandidaten ausgewertet wurden. Anhand dieser Daten konnte ich als erstes Hauptergebnis insgesamt mehr als 20 photometrisch bestätigte Neuendeckungen von entfernten röntgenhellen Galaxienhaufen bei z >0.9 nachweisen. Desweiteren hat es die neue Z–H Beobachtungsmethode∼ ermöglicht, erstmalig eine Galaxienhaufen-Stichprobe über die gesamte Rotverschiebungsspanne von 0.2 < z < 1.5 zu untersuchen. Durch den Vergleich der beobachteten Farbentwicklung der ‘Red-Sequence’∼ ∼ mit Entwicklungsmodellen konnte ich die Entstehungsepoche der Sternpopulationen der Frühtypgalaxien auf z = 4.2 1.1 einschränken und damit ihr hohes Alter bestätigen. Eine f ± weitere Vorstudie über die H-Band Leuchtkraftentwicklung von 63 hellsten Haufengalaxien (Brightest Cluster Galaxies, BCGs) über das gleiche Rotverschiebungsintervall ermöglichte es, erstmals direkte Beobachtungshinweise dafür zu liefern, dass sich die Masse der größten Gala- xien im lokalen Universum seit z 1.5 mindestens verdoppelt hat. Mein vorläufiges Ergebnis, ≃ dass nahe BCGs alte Sternpopulationen besitzen, ihre Masse aber erst zum Großteil über die letzten 9 Milliarden Jahre angesammelt wurde, stimmt qualitativ mit den Vorhersagen der mo- dernsten Simulationen zur hierarchischen Galaxienentstehung und -entwicklung innerhalb des theoretischen Standardmodells überein. Die Bestätigung und Verfeinerung dieser vorläufigen Ergebnisse werden zur Entwicklung einer widerspruchsfreien Beschreibung kosmischer Entwick- lung von Galaxienpopulationen und der großräumigen Struktur beitragen.

Studying Cosmic Evolution with the XMM-Newton Distant Cluster Project: X-ray Luminous Galaxy Clusters at z>1 and their Galaxy Populations ∼ Abstract Investigating X-ray luminous galaxy clusters at high redshift (z > 1) provides a challenging but fundamental constraint on evolutionary studies of the largest∼ virialized structures in the Universe, the baryonic matter component in form of the hot intracluster medium (ICM), their galaxy populations, and the effects of the mysterious Dark Energy. The main aim of this thesis work is to establish the observational foundation for the XMM-Newton Distant Cluster Project (XDCP). This new generation serendipitous X-ray survey is focused on the most distant galaxy clusters at z >1, based on the selection of extended X-ray sources, their identification as clusters and redshift∼ estimation via two-band imaging, and their final spectroscopic confirmation. As a first step, I have analyzed 80 deg2 (469 fields) of deep XMM-Newton archival X-ray data with a new pipeline processing system and selected almost 1 000 extended sources as galaxy cluster candidates, 75% of which could be identified as clusters or groups at z < 0.6 using available optical data. This left about 250 candidates with typical 0.5–2.0 keV X-ray∼ fluxes of 10−14 ergs−1cm−2 in need of confirmation as distant cluster sources. ∼ Therefore, I have adopted a new strategy to efficiently establish the nature of these extended X-ray sources and estimate their redshifts, based on medium deep Z- and H-band photometry and the observed Z–H ‘red-sequence’ color of early-type cluster galaxies. To fully exploit this technique, I have designed a new near-infrared data reduction code, which was applied to the data collected for 25% of the 250 distant cluster candidates in two imaging campaigns at the 3.5m telescope at the Calar Alto Observatory. As a first main result, more than 20 X-ray luminous clusters were discovered to lie at a photometric redshift of z >0.9. Furthermore, the new Z–H red-sequence method has allowed a cluster∼ sample study over an unprecedented redshift baseline of 0.2 < z < 1.5. From a comparison of the observed color evolution of the cluster red-sequence galaxies∼ ∼ with model predictions, I could constrain the formation epoch of the bulk of their stellar populations as z = 4.2 1.1. This confirms f ± the well-established old age of the stellar populations of early-type galaxies in clusters. The preliminary investigation of the H-band luminosity evolution of 63 brightest cluster galaxies (BCGs) over the same redshift range provides for the first time direct observational indications that the most massive cluster galaxies in the local Universe have doubled their stellar mass since z 1.5. My tentative finding that nearby BCGs have old, passively evolving stellar ≃ populations and were assembled in the last 9 Gyr is in qualitative agreement with predictions from the latest numerical simulations based on the standard cold dark matter scenario of galaxy formation and evolution via hierarchical merging. The confirmation and refinement of these preliminary results will contribute to the development of a consistent picture of the cosmic evolution of galaxy populations and the large-scale structure.

Contents

1 Introduction 1

2 Galaxy Clusters as Astrophysical Laboratories 3 2.1 GeneralProperties ...... 3 2.2 ClustersasDarkMatterHalos...... 4 2.3 X-ray Properties of the Intracluster Medium ...... 6 2.3.1 Emissionmechanisms...... 6 2.3.2 ICMstructure...... 8 2.3.3 ICMformation ...... 9 2.3.4 Scalingrelations...... 11 2.3.5 Coolcoreclusters...... 15 2.3.6 Sunyaev-Zeldovich effect ...... 16 2.4 TheClusterGalaxyPopulation ...... 17 2.4.1 Cluster versus field galaxies ...... 18 2.4.2 Formation of ellipticals and the cluster red-sequence...... 20 2.4.3 Evolution and environmental effects ...... 23 2.4.4 Brightestclustergalaxies...... 26

3 Galaxy Clusters as Cosmological Probes 29 3.1 CosmologicalFramework ...... 29 3.1.1 Dynamics of the expanding Universe ...... 29 3.1.2 Distancemeasures ...... 33 3.2 CosmicStructureFormation ...... 38 3.2.1 Growth and collapse of density perturbations ...... 38 3.2.2 Galaxyclusterstatistics ...... 41 3.3 Cosmological Tests with Galaxy Clusters ...... 51

4 X-ray Cluster Surveys 57 4.1 X-raySelection ...... 57 4.2 LocalClusterSurveys...... 58 4.3 DeepSurveys ...... 62 4.4 DistantGalaxyClustersin2004 ...... 64

i 5 The XMM-Newton Distant Cluster Project 67 5.1 Motivation...... 67 5.2 XDCPSurveyStrategy...... 68 5.3 PilotStudyResults...... 70 5.4 XMMUJ2235.3-2557 ...... 71 5.5 ThesisDataOverview ...... 75

6 X-ray Analysis of XMM-Newton Archival Data 77 6.1 XMM-NewtoninaNutshell ...... 77 6.2 X-ray Data Archive and Survey Definition ...... 83 6.3 Development of an X-ray Reduction and Analysis Pipeline ...... 85 6.3.1 Searchstrategy ...... 85 6.3.2 X-raydatareduction ...... 87 6.3.3 Detection of extended X-ray sources ...... 94 6.3.4 X-ray-opticaloverlays ...... 101 6.3.5 Distant cluster optimizations ...... 103 6.4 Source Screening and Candidate Selection ...... 104 6.4.1 Screeningsteps ...... 106 6.4.2 Classificationscheme ...... 108 6.4.3 DSSidentificationlimit...... 110 6.5 Performance Tests in the COSMOS Field ...... 112 6.6 FieldandSourceDiagnostics...... 114 6.6.1 XMM data yield and field statistics ...... 114 6.6.2 Extended source diagnostics ...... 117 6.7 Summary ...... 120

7 Near-Infrared Follow-up Imaging 121 7.1 ANewNIRImagingStrategy ...... 121 7.1.1 Motivation...... 121 7.1.2 Imagingstrategy ...... 123 7.1.3 Performanceexpectations ...... 126 7.2 PracticalImplementation...... 130 7.2.1 OMEGA2000inanutshell ...... 130 7.2.2 Observationalstrategy ...... 132 7.2.3 Imagingdataoverview ...... 133 7.3 Development of a NIR Science Reduction Pipeline ...... 134 7.3.1 Requirements and speciﬁcations ...... 134 7.3.2 NIRreductionsteps ...... 135 7.3.3 Faintobjectoptimization...... 143 7.3.4 Pipeline and instrument performance ...... 148 7.4 PhotometricAnalysis...... 153 7.4.1 Generalprocedure ...... 153 7.4.2 Photometriccalibration ...... 157 7.5 Summary ...... 161

8 Spectroscopic Follow-up 163 8.1 BasicsofSpectroscopy ...... 163 8.2 Challenges of Distant Galaxy Spectroscopy ...... 165 8.3 SpectroscopicDataReduction ...... 168 8.4 RedshiftDetermination...... 170 8.5 Summary ...... 172

9 XDCP Survey Status and Outlook 173 9.1 X-rayAnalysisStatus...... 173 9.1.1 Skycoverage ...... 173 9.1.2 Clusternumbercounts ...... 175 9.1.3 Selectionfunction...... 177 9.2 ImagingStatus ...... 180 9.3 SpectroscopyStatus ...... 183 9.4 XDCPOutlook ...... 183 9.5 Spectroscopically Conﬁrmed z>1Clustersin2007...... 185

10 High-Redshift Cluster Studies: First Results 187 10.1 The Formation Epoch of Red-Sequence Galaxies ...... 187 10.1.1 CalibrationoftheZ–Hmethod ...... 187 10.1.2 Discussion...... 190 10.2 Indications for Large-Scale Structure at z 1...... 191 ∼ 10.2.1 Observations ...... 191 10.2.2 Discussion...... 195 10.3 SummaryandConclusions ...... 200

11 Preliminary Studies and Science Outlook 201 11.1 The Assembly History of Brightest Cluster Galaxies ...... 201 11.1.1 BCGsampleandselection ...... 201 11.1.2 TheBCGH-bandHubblediagram ...... 203 11.1.3 Observing the BCG assembly epoch ...... 213 11.2 Cluster Red-Sequences Studies ...... 218 11.3 Gallery of High-Redshift Clusters ...... 219 11.4 SummaryandConclusions ...... 219

12 The Future of Galaxy Cluster Surveys 223 12.1SZESurveys...... 223 12.2eROSITA ...... 225 12.3 VADER–AConceptStudy ...... 226 12.3.1 Sciencecase ...... 226 12.3.2 Satellitedesignconcept...... 227 12.3.3 Surveyandresults ...... 228 12.4 ConcludingRemarks ...... 230

A XDCP Survey Field List 231

B List of Technical Terms and Acronyms 245 Chapter 1

Introduction

X-ray studies of clusters of galaxies have come a long way over the past four decades since the first (spurious) detection of X-ray emission from Coma and the speculations of its thermal origin (Felten et al., 1966). After the first pioneering imaging X-ray telescope EINSTEIN in the late seventies and the ground-breaking success of the ROSAT mission in the nineties, we are currently in the era of the third generation of X-ray observatories led by XMM-Newton and Chandra. Early survey results (Gioia et al., 1990; Henry et al., 1992) were suggestive that the population of X-ray luminous galaxy clusters exhibits a rather rapid evolution, which gave rise to the widespread belief that these objects were absent by redshifts of z 1. This ∼ was also predicted by theory for a Ωm = 1 Universe, the standard cosmological model at the time. Only one decade later, the situation had been completely reversed based on the results of the second generation cluster surveys. After discovering the first very distant systems, Rosati (2001) summarized the status for X-ray clusters in the following statement that has remained valid until the present: “ROSAT deep surveys have failed to show any dramatic change in the space density of clusters out to z 1, with the exception of the most luminous, massive systems.” ∼ Fundamental questions within the cosmic evolution scenario have so far lacked definite observational answers: (i) What is the formation epoch of the first X-ray clusters? (ii) How does the intracluster medium (ICM) form and when does it thermalize? (iii) When and how is the ICM polluted with metals and is it pre-heated? (iv) At what epoch do cluster galaxies form? (v) What drives their evolution in clusters? (vi) How and when do the most massive galaxies in cluster centers form? Great advances in observational cosmology over the last years have profoundly im- pacted the standard world model and have led to the establishment of a new ‘concordance’ cosmology. Galaxy clusters in this respect have provided early evidence of a low matter density Universe and have proven to be sensitive cosmological probes. Meanwhile, the properties and physical foundations of the Dark Energy component have emerged as the most fundamental open questions of the concordance model. One of the most promising approaches for constraining Dark Energy is provided by distant galaxy clusters. These open and central questions of cosmic evolution gave rise to the motivation to

1 2

initiate the XMM-Newton Distant Cluster Project (XDCP) as a third generation X-ray cluster survey with special focus on z> 1 systems. Starting about four years ago, the XDCP is a long-term, medium-size project with the following main goals:

1. Providing z > 1 galaxy clusters at various masses as individual astrophysical laboratories∼for detailed cluster and galaxy evolution studies (Chaps. 2 & 10); 2. Compiling pathﬁnder samples to characterize high-redshift clusters as cosmological probes for major upcoming SZE and X-ray surveys (Chaps. 2, 3, 9 & 12); 3. Establishing a statistically complete cosmological distant cluster sample selected from a suﬃciently large search volume for cosmological studies (Chaps. 3 & 9).

These different survey scopes are associated with increasing time lines. Whereas the current XDCP results mainly contribute to the first category, the project focus will shift towards the pathfinder role over the next 2–3 years and cosmological studies within about 4–5 years. The main scope for this work is to establish the observational foundation of the XMM-Newton Distant Cluster Project. The main agenda for the thesis can be subdivided into four parts:

Part I: The theoretical background and state-of-the-art in distant cluster studies is discussed in Chaps. 2–4. Chapter 2 introduces the basic properties of clusters, Chap. 3 establishes the cosmological framework and discusses clusters as cosmological probes, and Chap. 4 provides an overview of X-ray surveys. Owing to the rapid developments in the field over the last decade, the introductory concept aims at a broad overview of the current status in galaxy cluster research. Part II: The observational strategies and methods are introduced in Chaps. 5–8. Chapter 5 discusses the general XDCP survey strategy, Chap. 6 summarizes the X-ray analysis techniques, Chap. 7 reviews follow-up imaging methods, and Chap. 8 provides a brief overview of optical spectroscopy. The treatise of these technical chapters is aimed at providing a profound insight to the different observational methods to those without deep prior knowledge of the applied techniques. Part III: The survey status and first results are summarized in Chaps. 9–11. Chapter 9 evaluates the current status of the different survey stages and discusses open survey tasks. Chapter 10 presents first science applications based on spectroscopically confirmed clusters. Preliminary results of an extended photometrically characterized cluster sample are shown in Chap. 11. Part IV: The thesis closes with an outlook on future developments in Chap. 12. Chapter 2

Galaxy Clusters as Astrophysical Laboratories

This chapter will review selected aspects of the physical foundation of galaxy clusters. The main focus will be placed on the X-ray properties of the intracluster medium (ICM) and optical characteristics of the (distant) cluster galaxy population. The discussions on X-ray emission mechanisms, ICM structure and formation, and cool cores provide the background for understanding the high-redshift cluster detection procedure. X-ray scaling relations are critical for the mass calibration, whereas the red-sequence of cluster galaxies is used for the initial redshift estimation. The sections on galaxy formation, evolution, and environmental eﬀects set the stage for the high-redshift cluster studies in Chaps. 10 & 11. More details can be found in the classic work of Sarazin (1986), the review of Voit (2005), or Biviano (2000) for a historical treatise of clusters.

2.1 General Properties

From a modern point of view, galaxy clusters can be considered as dark matter halos, whose gravitational potential wells are filled with a dilute hot gas, the intracluster medium, and which confine the bound cluster galaxy populations. The mass fractions of these three dominant cluster components are approximately 80–85% for the dark matter (DM), 12– 15% for the intracluster medium, and only 2–5% for the total mass of the galaxies. Galaxy clusters occupy the top level of the cosmic object hierarchy and are extraordi- 14 15 nary objects in many respects. With typical masses of 10 –10 M⊙ and sizes of a few Mega parsecs (Mpc), they are by far the largest and most massive virialized objects in the Universe. Galaxy clusters also lead the object classes in terms of their dark matter domination, which is manifested in a mass-to-light ratio of M/L 200(M⊙/L⊙). On the other hand, clusters are unique for studying the baryonic matter≃ component, since they define the only large volumes in the Universe from which the majority of baryons emit detectable radiation. Clusters constitute important astrophysical laboratories for various fields in physics

3 4 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

and are observed and studied over the full electromagnetic spectrum. The dark matter component of the matter density is the main ingredient for cosmological structure formation theories in which the largest collapsed dark matter halos are identified with galaxy clusters (see Sect. 3.2). The theoretical dark matter distribution of clusters (Sect. 2.2) has been studied in detail with elaborate numerical simulation experiments for about three decades. New observational techniques using the weak gravitational lensing effect caused by the total mass concentration of a cluster have recently enabled direct observations of the DM structure of clusters. The hot intracluster medium (Sect. 2.3) can be considered as a giant plasma physics 3 2 5 −3 laboratory filling several Mpc of volume at very low electron densities of ne 10 –10 m , an environment which is not accessible for table top experiments. Owing≃ to the plasma 6 temperatures of kB T 2–10 keV, corresponding to T (20–100) 10 K, the thermal ICM ≃ 43 ≃ 45 × −1 emission gives rise to X-ray luminosities of LX 10 –3 10 erg s , manifesting clusters as the most X-ray luminous objects in the Universe≃ next× to Active Galactic Nuclei (AGN). X-ray observations of the intracluster medium of nearby clusters nowadays allow precision measurements of the ICM structure, its thermodynamic state, and elemental abundances. Furthermore, the ICM enables studies of hydrodynamical and plasma physical processes on the largest scales such as shock fronts, contact discontinuities, propagation of sound waves, turbulence, heat conduction, and diffusion processes. Last but not least, the galaxy populations of clusters (Sect. 2.4), accessible through optical and infrared observations, can be investigated in well defined environments for detailed studies on galaxy formation, evolution, and transformation processes. We will now have a closer look at the three major individual matter components of galaxy clusters, with a focus on important properties for high-redshift cluster studies.

2.2 Clusters as Dark Matter Halos

The missing mass problem was established 70 years ago in a seminal paper by Zwicky (1937), who noted that the visible mass in the Coma cluster was not consistent with total mass estimates based on the Virial theorem. Besides these very early indications for the need of a dark matter component, the best ‘direct’ empirical evidence for the existence of dark matter to date is also provided by galaxy clusters. Recent detailed studies of the ‘Bullet Cluster’ have shown that the center of mass of the luminous matter and the total mass distribution have been spatially separated as the consequence of a major merger event (Clowe et al., 2006). Figure 2.1 shows the observed X-ray emission of the cluster in pink and the weak lensing reconstructed total mass distribution in blue. The bullet-shaped ICM emission in the right part is trailing the total mass of the infalling subcluster after the initial collision. Being the dominant baryonic mass component, the ICM center of mass can only deviate from the total mass center if non-baryonic dark matter is the main cluster constituent.

1The image can be found at http://chandra.harvard.edu/photo/2006/1e0657/1e0657 scale.jpg. 2.2. CLUSTERS AS DARK MATTER HALOS 5

Figure 2.1: The ‘Bullet Cluster’ 1E 0657-56 at z =0.296 provides the best empirical evidence for the existence of dark matter (Clowe et al., 2006). The cluster was formed through a major merger event which spatially separated the cluster constituents. The X-ray emission of the ICM, as the dominant baryonic component, is shown in pink; the blue shade indicates the weak lensing reconstructed mass concentrations, which are coincident with the galaxy overdensities. The ﬂuid-like hot ICM gas trails the collisionless dark matter and galaxy components resulting in spatially distinct centers of the dominant visible matter (ICM) and the major mass constituent (dark matter). Image from press release1.

The properties of dark matter halos, i.e. DM aggregates that form a quasi equilibrium configuration, have been investigated in detail by means of high-resolution numerical simulations. Navarro, Frenk, and White (NFW) (1997) found a universal DM halo density profile for virialized objects for the now standard cold dark matter2 (CDM) paradigm. The widely used NFW fitting formula for the density profile ρDM(r) has the form

ρs ρDM(r)= 2 , (2.1) r 1+ r rs rs where rs is a characteristic scale length and ρs =δc ρcr(z) the central density (see Equ. 3.10), which is linked to the cosmological model (see Sect. 3.2). At small radii r rs, the NFW profile exhibits a ‘cusp’ associated with the inner slope ρ (r) r−1, whereas≪ at cluster- DM ∝ centric distances of r r the density follows ρ (r) r−3. ≫ s DM ∝ The characteristic central ‘cuspy’ shape of the NFW profile is a fundamental prediction of the CDM paradigm, quasi independent of halo mass and cosmological parameters. X-ray observations of galaxy clusters have meanwhile confirmed the predicted universal CDM profile (e.g. Pratt & Arnaud, 2002; Pointecouteau et al., 2005), as shown in Fig. 2.3.

2Cold means that the kinetic energy of the particles is much smaller than their rest mass energy. 6 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

Figure 2.2: X-ray emission of the ICM. Left: Unabsorbed ICM model spectra of the bremsstrahlung continuum and line emission for plasmas with 0.4 solar metallicity and temperatures of 1 keV (black), 3 keV (red), and 9 keV (green). Right: Absorbed model spectra of T =3 keV observed through galactic hydrogen columns of 1021 cm−2 (green), 3 1020 cm−2 (red), and for the unabsorbed case (black). × Plots from Schneider (2006a).

Concerning clusters at high redshift, the NFW proﬁle predicts denser cores compared to recently formed objects as a consequence of the increased critical density ρcr(z) at collapse time (see Sect. 3.2).

2.3 X-ray Properties of the Intracluster Medium

“Clusters of galaxies are X-ray sources because galaxy formation is ineﬃcient” (Voit, 2005). About 90% of all baryons in the Universe reside in intergalactic space, but X-ray emitting gas temperatures are only reached in the deep potential wells of galaxy clusters. The main X-ray properties of the hot intracluster medium will be summarized in this section.

2.3.1 Emission mechanisms

The dominant X-ray emission mechanism for X-ray clusters at temperatures of T > 107 K ∼ (kBT > 1 keV) is thermal bremsstrahlung (free-free radiation). The intracluster medium is a collisionally∼ ionized, optically thin plasma, which implies that essentially every emitted ﬀ photon will escape from the cluster volume. The bremsstrahlung emissivity ǫν for a plasma at temperature T , i.e. the luminosity per unit volume and frequency interval, is given by

hP ν − k T ﬀ −38 2 ﬀ e B −1 −3 −1 ǫ 6.8 10 Z ne ni g (ν, T ) erg s cm Hz , (2.2) ν ≈ × i √T 2.3. X-RAY PROPERTIES OF THE INTRACLUSTER MEDIUM 7

3 where ne and ni denote the number densities of electrons and ions, Zi the ion charge, and gff (ν, T ) ln(9k T/(4h ν)) is the Gaunt factor, which is of order unity and corrects ∝ B P for quantum mechanical effects and distant collisions. Equation 2.2 implies that thermal bremsstrahlung spectra are essentially constant (in frequency space) for h ν k T and P ≪ B exhibit a temperature dependent steep exponential cut-off at high frequencies of hPν >kB T . One of the key features of X-ray observations is the proportionality of the bremsstrahlung∼ emissivity (and all other radiation processes induced by particle collisions) to the square of the gas density ǫff n n n2, which results in a peaked, high contrast signature of the ν ∝ e i ≈ e central cluster regions. The total integrated (bolometric) volume emissivity of a thermal plasma with solar metal abundances is then

∞ T n 2 ǫﬀ = dν ǫﬀ 9.5 10−32 e erg s−1cm−3 √T . (2.3) bol ν s −3 −3 Z0 ≈ × 1 keV 10 cm ∝ At temperatures of T < 2 keV, recombination radiation (free-bound) and line emission radiation (bound-bound)∼ become important4 as the ionization levels of the ICM metals decrease. For ions with an ionization potential χi of state i, the free-bound emissivity has the form

− hP ν χi − k T fb fb e B ǫν g (ν, T ) 3 . (2.4) ∝ T 2 The most important X-ray line feature for massive clusters is the K-shell line complex of hydrogen-like iron Fe XXVI around 6.7 keV, with slightly shifted energies for other ionization states. At lower temperatures, additional important line features originate from the Fe L line complex at 1 keV and ions of O, Mg, Si, S, Ar, Ca, and Ne. ∼ Figure 2.2 illustrates in the left panel three thermal ICM model spectra for plasma temperatures of 9 keV (green), 3 keV (red), and 1 keV (black) including the bremsstrahlung continuum, recombination radiation, and line emission. The right panel shows the effects of the interstellar medium (ISM) absorption on an observed 3 keV cluster spectrum for hydrogen column densities of 1021 cm−2 (green), 3 1020 cm−2 (red), and in the unabsorbed × case (black). The photoionization cross-section of neutral hydrogen, and similarly for the heavier elements of the ISM, exhibits approximately the photon energy dependence a E−3 (Cox, 2000), i.e. the extinction increases rapidly with decreasing energy (above E ∝ the absorption edge). As can be seen from Fig. 2.2, the galactic ISM extinction introduces an effective cut-off in the observed spectrum at energies of about <0.3 keV for the typical hydrogen columns of 2–5 1020 cm−2 in extragalactic survey fields.∼ × The total luminosity of a cluster is obtained from the emissivity by integrating over the plasma volume and frequency L = ǫ dV dν EM, where EM = n2 dV is X ν V ν ∝ V e the emission measure. Observed along theR line-ofR sight at projected radius r =R dang(z)Θ 3To avoid confusion with the Hubble constant parameter h and wavenumbers k, the Boltzmann constant is denoted as kB and the Planck constant as hP throughout this thesis. 4Modern ICM radiation codes account for all emission mechanisms at all temperatures. 8 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

2 4 (see Equ. 3.21), the emission measure EM = ne dl =4 π (1+z) S(Θ)/Λ(T, z) gives direct observational access to the local electron densityR ne, deduced from the measured surface brightness proﬁle S(Θ) as a function of cluster-centric angle Θ in the soft X-ray band (see Equ. 2.6), and the redshifted and absorption corrected emissivity Λ(T, z). The gas temperature TX, as the second local X-ray observable, is obtained from a model ﬁt to the redshifted X-ray spectrum, which achieves the highest temperature sensitivity in the hard band region.

2.3.2 ICM structure The angle-averaged global spatial structure of the intracluster medium is commonly described by the King model for a self-gravitating, isothermal sphere (King, 1966). If one allows for an additional scaling of the ICM density profile with respect to the underlying dark matter profile of the form ρ ρβ , the three dimensional β-model for the radial gas ∝ DM ICM profile ρgas(r) is obtained (Cavaliere and Fusco-Femiano, 1976)

ρgas,0 ρgas(r)= 3β . (2.5) 2 2 1+ r rc The core radius rc determines the characteristic extent scale of the source, whereas the β parameter modulates the overall steepness of the profile. The original King model for the underlying dark matter potential is recovered by setting β =1. Note that the central gas density ρgas,0 defines the peak of the core of the profile, in contrast to the rising cusp of the NFW profile of Equ. 2.1. The β-model for the global ICM structure assumes a spherically symmetric, isothermal gas in hydrostatic equilibrium. For β parameters deviating from unity, the galaxies, as dynamic tracers of the dark matter potential, exhibit a different velocity dispersion than 2 2 the ICM gas according to σgal =β σgas. Typical values of β 2/3 0.2 (Jones and Forman, 1984) indicate that the galaxy velocity dispersion in clusters≃ is lower± than for the gas. In effect, β is the ratio between the kinetic energy of tracers of the gravitational potential 2 (e.g. galaxies) and the thermal energy of the gas β = µ mp σr /(kBTgas), where σr denotes the radial velocity dispersion, mp the proton mass, and µ the mean molecular weight of the gas (µ 0.6 for a primordial gas composition). The empirically determined β values imply that the≃ mean energy per unit mass is higher for the intracluster medium component than for the galaxies. The observed two dimensional X-ray surface brightness profile S(Θ) of the β-model is obtained by relating the gas density profile ρgas(r) to the X-ray emissivity (Equ. 2.2) and projecting the distribution on the plane of the sky. The resulting radially symmetric profile

S S 2 S(Θ) = 0 = 0 for β = (2.6) 3·β− 1 3 2 2 2 2 3 1+ Θ 1+ Θ Θc Θc 2.3. X-RAY PROPERTIES OF THE INTRACLUSTER MEDIUM 9

contains the fit parameters Θc for the angular core radius size, the central surface brightness S0, and the β value. In the last step, the slope parameter is fixed to β =2/3 in order to reduce the number of degrees of freedom for model fits of weak sources. This constrained surface brightness profile serves as a reference model for the detection of serendipitous extended X-ray sources (e.g. clusters) in Sect. 6.3.3, which contain typically a few hundred X-ray photons. Detailed investigations of the morphological properties of nearby clusters (e.g. Böhringer et al., 2007b) require more elaborated structural parameters that account for multiple components and deviations from spherical symmetry.

2.3.3 ICM formation Studies of the ICM formation process will be one of the important future applications of the XDCP survey; some basic considerations and models are now discussed. Scenarios for the intracluster medium formation need to account for two essential observational constraints: (i) the total ICM mass of M 1014 M for massive clusters, and (ii) the metallicity ICM ∼ ⊙ enrichment of the intracluster gas with values of ZICM 0.4 Z⊙. We will first take a look at infall models for the ICM,∼ which were proposed shortly after the first X-ray observations of clusters became available (e.g. Gott&Gunn, 1971; Gunn & Gott, 1972). Let us consider a primordial, cool gas initially at rest at large distance that falls into the potential well of an existing massive dark matter halo. Applying the Virial theorem for the final equilibrium state of the gas settled to the bottom of the potential, we obtain

2 G M 3 kBTgas G M 2 E¯kin + E¯pot =0 v , (2.7) ≃h i − Rg ≃ µ mp − Rg 2 2 2 where v σ =3 σr is the squared (radial) velocity dispersion of particles in the potential h i≡ 2 and Rg G M / E¯pot is the gravitational radius of the cluster. In the last step, the kinetic particle≡ energy was| equated| with the thermal energy of a monoatomic gas (i.e. β 1). This ﬁrst order approximation yields ICM gas temperature estimates expressed in terms≃ of the velocity dispersion of the potential of

µ m σ2 σ 2 σ 2 T p r 7.3 107 K r 6.3 keV r (2.8) gas 3 −1 3 −1 ≃ kB ≃ × · 10 km s ≃ · 10 km s in reasonably good agreement with observations. The hot ICM temperatures can hence be explained as a natural consequence of gas infall to the dark matter halo. The total ICM mass can be accounted for when considering that clusters form via collapse of cosmic regions of Rcol (10 Mpc) with an associated approximate baryon reservoir of Mbar (4/3) π R3 ρ ∼Ω O(1+δ) 1014 M (see Sect. 3.2 for explanation). ∼ col · cr b ∼ ⊙ Several assumptions for the derived temperature estimate might not be well justiﬁed and could modify the gas temperatures: (i) the ICM infall after the DM halo formed, (ii) initially only marginally bound gas coming from a large distance, (iii) a cold medium before infall, and (iv) no radiative energy losses during collapse. 10 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

The purely gravitationally driven thermalization process of the ICM can be qualitatively described with a spherically symmetric smooth accretion model: (i) incoming cold gas enters the cluster in concentric shells, (ii) the gas is adiabatically compressed as it reaches the denser central regions, (iii) accretion shocks form when the gas motions become su- personic and the shocks propagate outwards, (iv) subsequent infalling shells pass through the accretion shocks of the inner shells and are shock-heated to the virial temperature of the cluster within about one sound crossing time tsound. For hierarchical merger scenarios, accretion shocks are not well deﬁned but form a complex network of shocks as diﬀerent infalling subclusters mix with the ICM of the main halo. After shocks have passed, the ICM gas is in a nearly hydrostatic state and further evolution proceeds quasi-static. Since the cluster and ICM formation physics contains no characteristic scale, a self-similarity of the dark matter and gas internal structure is expected, i.e. the main object properties should only be scaled versions of each other. The characteristic cluster formation time scale is given by the sound crossing time tsound which can be expressed as

− 1 R R Tgas 2 tsound 0.66 Gyr 8 (2.9) ≃ csound ≃ · 1 Mpc! 10 K derived from the intracluster medium sound speed of

1 2 −1 Tgas csound 1480 kms 8 . (2.10) ≃ · 10 K The sound crossing timescale is hence short compared to the ages of low-redshift clusters, implying that the central cluster gas will reach a quasi-hydrostatic equilibrium about 1 Gyr after formation or major merger events. The situation for clusters at redshift z 1.5 with a cosmic age of 4 Gyrs (see Fig. 3.3) is already quite diﬀerent and one would∼ expect to ∼ ﬁnd a large fraction of systems that have not yet fully virialized. Equation 2.9 additionally implies that cosmic regions with dimensions of >10 Mpc will not be able to virialize within a Hubble time and hence no relaxed objects larger∼ than the cluster scale exist at the current epoch. The ICM sound speed (Equ. 2.10) is comparable to the velocities of the cluster galaxies. The galaxy crossing time scale for a passage through the cluster tcross

−1 R R σv tcross 1Gyr 3 −1 (2.11) ≃ σv ≃ · 1 Mpc! 10 km s

is hence similar to tsound implying that the same arguments apply for the relaxation time scale of the galaxy population. ICM infall models can reproduce basic properties of the intracluster medium, but do not account for the observed heavy element abundances of the initially metal free primordial gas. The detection of the 6.7 keV iron line (Mitchell and Culhane, 1977) provided early evidence that at least some of the gas must have been processed by stars and later been ejected from the galaxies. Heavy element abundances and their distribution in low 2.3. X-RAY PROPERTIES OF THE INTRACLUSTER MEDIUM 11

redshift clusters can now be consistently attributed to supernova explosion yields, which also provide sufficient energy for ejecting the metals out of the galaxies into the ICM (e.g. Böhringer, 2003). As a last point concerning the formation process, we can now address the question of what determines the density of the intracluster medium, which is closely linked to the 5 2/3 entropy of the ICM. The cluster entropy K kB T/(µ mp ρgas ) and the shape of the dark matter potential well are the main determinants≡ for the ICM X-ray structure (Voit, 2005). As a second fundamental key feature, the cluster entropy contains a record of the thermodynamic history of the ICM gas. The ICM structure of clusters is influenced by K because, in a simple picture, high entropy gas floats and low entropy gas sinks. The ICM gas thus convects6 until the equilibrium condition dK/dr 0 is fulfilled throughout the cluster, i.e. a quasi-static configuration is achieved when the isentropic≥ surfaces coincide with the equipotential surfaces of the dark matter potential. For a purely gravitationally driven ICM collapse with initially cold gas, the accretion shocks are the only source of entropy for the ICM gas, with the consequence that the self-similar entropy profile K(r) r1.1 depends entirely on the mass accretion history ∝ M(t). In the case that the accreted gas was not cold, an isentropic central entropy core is expected with a level similar to the pre-shock entropy. This so called pre-heating of ICM gas before infall makes gas harder to compress and thus lowers the density in the cluster center, where entropy is smallest. This mechanism can give rise to a breaking of the cluster self-similarity and is hence of great interest. Observations seem to be consistent with a minimum entropy floor7 level of about 100–200 keV cm2, but the simple pre-heating model is likely to be complemented by entropy changing radiative cooling and heating processes. Many fundamental questions concerning the intracluster medium formation have remained unanswered and are awaiting investigations using appropriate high-redshift X-ray cluster samples. (i) At what epoch does the thermal ICM X-ray emission start? (ii) Are non-gravitational processes important during the formation process? (iii) Is the gas pre- heated before falling into the cluster potential well? (iv) What is the source of (extra) entropy? (v) What is the early metal enrichment history and thermal evolution of the ICM?

2.3.4 Scaling relations The expected (nearly) self-similar cluster formation process gives rise to general scaling relations of physical cluster properties. The comparison of observed relations with the

5 5/3 K is the constant of proportionality in the equation-of-state for an adiabatic monoatomic gas P =Kρgas 3/2 and is related to the standard thermodynamic entropy per particle through s = kB ln K + s0. An 2/3 alternative, widely used deﬁnition for the cluster entropy is S =Ke =kB T/ne . 6Observations deviate from the expected polytropic index of γ =5/3 for a convectively established equilibrium, suggesting that also other processes are signiﬁcant. 7This picture has recently been adjusted in terms of a gradual change of the entropy level as a function of ICM temperature. 12 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

Figure 2.3: Galaxy cluster mass profiles for nearby clusters. Left: Integrated total mass profiles as a function of physical radius (kpc) for ten clusters with temperatures in the range 2–9 keV. Solid lines represent the best fitting NFW profiles. Right: Scaled mass profiles for the same clusters with masses scaled to M200 and radii to R200. The black line gives the mean scaled NFW model. Plots from Pointecouteau et al. (2005).

scaling behavior of the self-similar predictions yield important clues on additional contributing physical processes. Scaling relations for distant cluster applications provide the crucial mass proxy in absence of high quality X-ray data for a detailed analysis, which requires > 1 000 X-ray photons. We will first introduce the standard method for deriving total cluster∼ masses from X-ray data and will then discuss some important scaling relations. With XMM-Newton and Chandra, the following physical cluster properties can be directly derived from sufficiently deep X-ray data: (i) the X-ray luminosity LX from the source flux fX (and redshift), (ii) the radial gas density profile ρgas(r) from the local emission measure EM(r) (iii) the total gas mass Mgas from the integrated density profile, (iv) the global mean temperature TX from the observed spectrum, and (v) the temperature profile TX(r) from spatially resolved spectra. Using the spatially resolved density and temperature distributions, the (vi) entropy profile K(r) and (vii) the total integrated mass profile Mtot(< r) can be obtained. Under the assumptions of spherical symmetry and hydrostatic equilibrium, the integrated total mass profile Mtot(< r) can be derived from the ideal gas equation pgas = ρgas kB TX/(µ mp) and the hydrostatic equation for pressure equilibrium

dpgas(r) G Mtot(

Solving for Mtot(

Figure 2.4: Scaled temperature pro- ﬁles of 15 nearby clusters. The grey shaded area indicates the region enclos- ing the 1 σ deviations around the mean proﬁle. Temperatures are scaled to the averaged temperature in the outer cluster parts. Plot from Pratt et al. (2007).

kB T (r) r d ln ρgas d ln T Mtot(

M200 4 π 3 = ρ vir = δ ρcr(z) , δ 200 . (2.14) 3 R200 h i ∼ The left panel of Fig. 2.3 shows the determined integrated total mass profiles of ten nearby clusters within the temperature range of 2–9 keV (Pointecouteau et al., 2005) with the best- fitting NFW dark matter profiles (Equ. 2.1) indicated by the solid lines. Once the mass profiles are scaled to units of R200 and M200 (right panel) the clusters reveal their self-similar property and are all consistent with a single universal NFW mass profile illustrated by the solid line. A similar universal scaling behavior can be observed for the outer regions of the temperature profiles as displayed in Fig. 2.4 (Pratt et al., 2007). The inner cluster regions at R <0.1 R200 exhibit an increased temperature scatter due to additional non-gravitational physics∼ in cluster centers (Sect. 2.3.5). 14 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

As for the scaled radial proﬁles, a self-similar behavior is expected for global properties (Kaiser, 1986). Under the assumptions that for systems in virial equilibrium (i) ﬀ the bremsstrahlung emissivity scales with temperature as ǫbol √TX (Equ. 2.3), (ii) the 3 ∝ mass scales with a characteristic radius R∗ as Mtot R∗, (iii) a constant gas mass fraction f =M /M const, and (iv) self-similar dark matter∝ halos (Equ. 2.1), global relations gas gas tot ≈ for purely gravitationally driven cluster physics can be derived. Since the NFW dark matter halos form a two parameter family, where the parameters are the total halo mass Mtot and the formation epoch zform, the relations are expected to hold for all clusters that formed at the same redshift. The proper redshift dependence of the scaling relations can additionally be accounted for by considering the redshift evolution of the collapse threshold (Kitayama and Suto, 1996), which predominantly depends on the critical density at a given redshift 2 ρcr(z)=E (z) ρcr(0), parameterized by the evolution function E(z) (see Equs. 3.10 & 3.15), and to a lesser extent on the cosmological model parameters (see Sect. 3.2). As an example for the full redshift dependence, we consider the evolution of the cluster size scaling relation R (z) M 1/3 E−2/3(z), which indicates8 that objects at high redshift have smaller 200 ∝ 200 dimensions and are hence more compact. We will now return to the discussion of four important global scaling relations for clusters of the same formation epoch. A summary of these relations is listed in Tab. 2.1. For a more detailed discussion of empirical parameters see Zhang (2005).

2 2/3 3/2 M–TX: Equation 2.7 yields T σ M/R M and consequently M T . Slopes of X ∝ r ∝ ∗ ∝ ∝ X empirical M–TX relations are consistent with the self-similar expectation(e.g. Arnaud et al., 2005). The M–TX relation is one of the fundamental scaling laws because of the direct connection of mass and gas temperature through the Virial theorem. With the assumed constant gas mass fraction Mgas = fgasM, the same self-similar scaling behavior is obtained for the Mgas–TX relation.

bol bol LX –TX: An averaged integration of the emissivity over the cluster volume yields LX 2 3 1/2 3 3 1/2 1/2 2 ∝ ρgas R∗ TX (fgasM/R∗) R∗ TX M TX TX, where in the last step the result for ∝ ∝ ∝ bol the M–TX relation was used. The measured scaling behavior of the LX –TX relation reveals a systematically steeper logarithmic slope than expected.

Lbol–M: Combining the ﬁrst two relations, we obtain Lbol T 2 M 4/3. The Lbol–M cor- X X ∝ X ∝ X relation is of fundamental importance for distant cluster cosmology since it provides the easiest means to determine masses for faint X-ray clusters. Reiprich & B¨ohringer (2002) empirically calibrated the relation and obtained a slightly steeper slope than the self-similar expectation (see Fig. 4.2). The approximate full empirical relationship between X-ray luminosity and total cluster mass is given by

M 1.8 L 1045.0±0.3 h−2 erg s−1 200 . (2.15) X 70 15 −1 ≃ 10 h70 M⊙ !

8For E(z) 1, e.g. for a concordance model. ≥ 2.3. X-RAY PROPERTIES OF THE INTRACLUSTER MEDIUM 15

Relation Form Self-Similar α Observed α Reference α M500–TX M500 TX 3/2 1.6 0.1 Arnaud et al. (2005) bol bol ∝ α ± LX –TX LX TX 2 2.9 0.15 Arnaud&Evrard (1999) bol bol ∝ α ± LX –M200 LX M200 4/3 1.8 0.1 Reiprich & B¨ohringer (2002) M –Y M ∝ Y α 3/5 0.55 ± 0.05 Arnaud et al. (2007) 500 X 500 ∝ X ± Table 2.1: Selected galaxy cluster scaling relations. The table lists the relation, the expected self-similar scaling behavior, and approximate observed values.

M–YX: The use of the combined parameter Y = M T as a robust, low-scatter mass X gas · X proxy has been motivated by studies of the Sunyaev-Zeldovich effect (see Sect. 2.3.6) and has recently been calibrated with X-ray observations by Arnaud et al. (2007). The self-similar expectation is obtained from the first relation as Y M T X ∝ gas X ∝ f M M 2/3 M 5/3 and hence M Y 3/5, consistent with the observed slopes. gas ∝ ∝ X The introduced scaling relations are of fundamental importance for the XDCP survey in two ways. First, a selected subset of newly discovered individual distant systems can be used to study and calibrate the evolution of the mass-observable relations at high redshift, possibly with deep X-ray follow-up observations in order to improve the achievable accuracies. Second, the calibrated scaling relations can then provide the mass proxy for future cosmological applications for the full survey sample. Deviations from the expected gravitational self-similar scaling relations are commonly attributed to non-gravitational effects in the intracluster medium. ICM pre-heating and central cluster cooling and heating processes have been mentioned as possible mechanisms able to break self-similarity. The next section will briefly summarize some of the complex baryonic physics phenomena observed in cluster centers.

2.3.5 Cool core clusters Cooling and heating in the centers of galaxy clusters is the subject of very active current research since the observed physical processes affect the evolution of both the cluster and the galaxy populations. A compilation of the latest results is available in the conference proceedings of Böhringer et al. (2007a). Cool core clusters (CCC) could also be important for high-redshift studies for two reasons. (i) The abundance and properties of high redshift cool core clusters are essentially unexplored (e.g. Vikhlinin et al., 2007; Santos et al., submitted), and (ii) CCC systems at z> 1, if they exist, could introduce survey selection biases owing to their highly peaked X-ray emission and the associated difficulty to discriminate them from point sources. The radiative cooling time scale of the ICM can be approximated from the ratio of the specific enthalpy9 and the total volume emissivity and is inversely proportional to the gas density n

9The speciﬁc enthalpy h is to be used for an isobaric cooling of the gas. 16 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

5 −1 1/2 2 nkB T 10 n T 1 tcool ff 8.5 10 yr −3 −3 8 . (2.16) ≃ ǫbol ≃ × · 10 cm 10 K ∝ n Throughout most regions of typical clusters, the cooling time scale is much longer than the −1 Hubble time tH = H0 , justifying the assumption of hydrostatic equilibrium. However, in dense cluster cores, tcool can become less than the age of the Universe and radiative cooling effects gain significance. The cooling surface is defined as the radius rcool, where the cooling time equals the age of the cluster. Inside rcool radiative cooling losses are significant and can cause a decline of the ICM temperature towards the center, as visible in some of the temperature profiles of Fig. 2.4. Such cool core clusters exhibit very peaked surface brightness profiles with central emission exceeding the β-model expectations of Equ. 2.6. Contributing about half to the local cluster population, current data is suggestive that the fraction of CCC is rapidly declining at z >0.5 (Vikhlinin et al., 2007). Early models of cool core clusters (e.g.∼ Silk, 1976; Fabian & Nulsen, 1977) predicted that the central cooled ICM should gradually condense and be replaced by surrounding gas giving rise to a constant inward flow with mass deposition rates in the cluster centers of (100 M ) per year. However, observations with XMM-Newton showed that this cooling O ⊙ flow picture does not hold (Peterson et al., 2001) and that the central gas temperatures do not decrease below a few keV. These results emphasized the need of a central heating source, which is sufficiently energetic and self-regulated to balance the radiative cooling losses and establish a stable ICM configuration. The current standard paradigm (e.g. Böhringer et al., 2002b) attributes the heat source to a central AGN, which exhibits duty-cycles of increased activity and intermediate quiescent periods. This paradigm is consistent with observations of complex hydrodynamic phenomena in cluster centers such as X-ray cavities and shock fronts (e.g. McNamara et al., 2005), buoyantly rising bubbles and propagating sound waves (e.g. Churazov et al., 2001; Fabian et al., 2003), or AGN jet interactions (e.g. McNamara et al., 2007). For cooling time scales of (1 Gyr), characteristic features of CCC might already be observable at z> 1. The associatedO peaked X-ray surface brightness profile and possible additional central AGN activity have to be adequately considered for the cluster candidate selection process (Sect. 6.4) and the computation of the selection function (Sect. 9.1.3).

2.3.6 Sunyaev-Zeldovich effect Measurements of the Sunyaev-Zeldovich effect (SZE) are expected to have a significant impact for future distant cluster studies and are therefore briefly summarized here. The hot electrons of the intracluster medium not only emit X-ray photons, they also lead to a characteristic distortion of the cosmic microwave background (CMB) radiation through inverse Compton scattering. Low energy CMB photons propagating through the hot cluster ICM are on average ‘upscattered’ by the electrons, i.e. the collective effect of many scattering processes gives rise to a shift of the CMB spectrum towards higher temperatures. The SZE (Sunyaev and 2.4. THE CLUSTER GALAXY POPULATION 17

−4 Zeldovich, 1972) introduces spectral changes of the CMB intensity I of order ∆I/I0 10 resulting in an increased CMB intensity in the higher energy Wien tail of the spectrum∼ and a CMB decrement in the Rayleigh-Jeans tail. The transition from a CMB decrement to an increment occurs at a wavelength of λnull 0.14cm ( 218 GHz). The SZE hence imprints a characteristic CMB spectral modulation≃ of a decr∼eased CMB intensity at λ> λnull (ν<218 GHz), a null effect at λnull, and an increment at λ<λnull (ν>218 GHz). The CMB decrement in the radio regime can be quantified as ∆I(ν)= 2 y I(ν) by defining the Compton-y parameter −

σT kB y 2 Te ne dl , (2.17) ≡ me c Z where σT is the Thomson electron cross-section, me the electron mass, and the integration runs along the line-of-sight path element dl. To ﬁrst order, the distorted CMB spectrum depends on the single y parameter, which is proportional to the probability that a photon will Compton scatter and the amount of energy the photon gains on average. A comparison to the ideal gas law p=nkB T reveals that the y parameter is proportional to the integrated pressure along the line-of-sight y p¯. The linear dependence on the gas density y ne implies that cluster outskirts can∝ be probed and that the SZE signal is less sensitive∝ to ICM inhomogeneities compared to X-ray measurements. In the case that the SZE signal of the cluster is not spatially resolved, an integrated version of the distortion parameter Y over the projected surface area is measured, Y = y dA n T dV E , which is proportional to the total thermal energy of the ∝ e ∝ thermal electrons.R Since the Y parameter is directly linked to the cluster gas mass and the temperature, a scaling relation of the form M Y α promises to yield a new robust mass proxy tot ∝ once calibrated. The key feature of the SZE concerning distant cluster studies is its (almost) redshift independence (for resolved systems). The cosmological surface brightness dimming Sbol (1+z)−4 (see Equ. 3.22) of the SZE signal is exactly compensated by the increased CMB∝ 4 4 intensity at the given redshift ∆I I TCMB (1+z) . For this reason, future SZE cluster surveys (see Sect. 12.1) bear great∝ hope∝ for a∝ profound impact on cosmological studies.

2.4 The Cluster Galaxy Population

Despite the small contribution of only a few percent to the total cluster mass, optical observations of the galaxies have long been the only means for cluster studies. Abell (1958) compiled the ﬁrst large optically selected cluster catalog based on galaxy overdensities ′ within a projected radius of ΘA =1.7/z, corresponding to a physical scale of about 2 Mpc. Redshift estimates for the 1 682 identiﬁed clusters containing more than 50 (projected) galaxies within two magnitudes of the third brightest member m3 were obtained by using the tenth brightest cluster member m10 as a ‘standard candle’ (see Equ. 3.46). Galaxy clusters play an important role for observational studies of galaxy evolution for several reasons: (i) they provide a large number of galaxies with the same general 18 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

environmental parameters and redshift, (ii) evolutionary processes are accelerated in high- density environments, and (iii) some classes of galaxies are only found in clusters. By going to high redshift, galaxy evolution effects are directly accessible to observations. Future studies of the evolutionary effects on the cluster galaxy populations will explore galaxy properties as a function of redshift and cluster mass over an increased redshift baseline. Whereas the cluster redshift fixes the cosmic age at the time of observation, the mass of the system determines the environmental properties. Disentangling the influence of different cluster specific effects on galaxies will thus require also a wide mass range at high redshift. The following sections illustrate selected aspects of the cluster galaxy populations that are of importance for distant cluster surveys.

2.4.1 Cluster versus field galaxies The galaxy population is commonly divided into cluster galaxies and field galaxies depending on the environment they reside in. In the light of hierarchical structure formation scenarios, this distinction is not well defined over cosmic time as galaxies from the ‘field’ can enter cluster or group environments and groups subsequently merge with larger clusters. As a working definition, we can consider as cluster galaxies all objects which are located within the virial radius of a massive dark matter halo filled with an X-ray emitting 43 −1 ICM of minimum luminosity LX > 10 erg s . Based on this ‘cluster definition’, three main differences distinguish the cluster∼ environments from the ‘field’:

the conﬁning dark matter potential well associated with the DM density proﬁle • of Equ. 2.1;

the dilute hot intracluster medium permeating the potential well with increasing • densities towards the center;

the highly increased galaxy density following a similar radial profile to the ICM. • All items of this special environment, as known from existing clusters, can introduce nur- ture effects on the galaxy population, i.e. object properties are transformed as a reaction to the environment. On the other hand, differences in the cluster versus field galaxy populations could originate from their different nature, i.e. they are inherited from the structure formation process covered in Sect. 3.2. In the following, some major consequences of the different physical cluster setting on the galaxy population are discussed.

Violent relaxation: A fluctuating gravitational potential Φ, as expected during the cluster collapse process, changes the energy of particles (galaxies or DM) according to E˙ = m Φ˙ and can produce a thermal equilibrium configuration on short timescales even for collisionless systems (Lynden-Bell, 1967). The initial cluster galaxy population is hence expected to be virialized shortly after cluster formation due to the collective gravitational effects of violent relaxation. 2.4. THE CLUSTER GALAXY POPULATION 19

Dynamical friction and cannibalism: A massive galaxy moving at velocity v through a homogeneous background medium of lighter (dark matter) particles with mass density ρ suﬀers a drag force named dynamical friction (Chandrasekhar, 1943). Along the trajectory of the massive object, the background medium is ‘gravitationally po- larized’ by the attractive force of the galaxy, leaving an overdense concentration of background particles in the wake of the galaxy. The collective eﬀect of this friction force results in a slowing of the galaxy according tov ˙ ρ M v−2. The drag force is independent of the mass of the individual background∝ particles, but proportional to the galaxy mass M, which leads to a mass segregation of the largest galaxies in clusters on the dynamical friction time scale tDF

−1 3 2 −2 M v R σr tDF 2Gyr 12 −1 −1 , (2.18) ≃ 10 M⊙ ! 1000kms 100 kpc! 300 km s

where R is the distance from the cluster center, and σr the radial velocity dispersion of the dark matter halo potential well. Massive cluster galaxies can hence spiral towards the center on time scales of a few Giga years and form a giant object via a process named galactic cannibalism, the growing of a single galaxy by consuming its neighbors (Hausman and Ostriker, 1978).

Galaxy merging: The key parameter to determine the probability that two approaching galaxies will merge is their relative velocity V , which needs to be smaller than the relative escape velocity of the system of (200kms−1) for a merging event to occur. If a signiﬁcant fraction of the orbital energyO is transferred to the stellar systems by tidal forces in a close encounter, a merger is likely. For two equal mass, spherically symmetric galaxies with a mass-weighted mean-square radius r2 and relative velocity V , merging occurs if the bodies have an impact parameterh ib smaller than a 2 2 2 1/4 critical value bcrit =(32 G M r /3) /V √M/V (Peacock, 1999). The expected merger rate Γ can then be approximatedh i as∝ the product of the merging cross section 2 σmerge =π bcrit and the particle ﬂux jgal =ngalV

1 Γ= σ j π b2 n V . (2.19) merge gal ≃ crit gal ∝ V The inverse proportionality to the velocity generally favors merger events in lower mass groups rather than in clusters, where velocity dispersions are higher.

Galaxy harassment: Even if galaxies do not merge, the cumulative effects of the tidal forces of many weaker encounters can significantly influence cluster galaxies, in particular objects with lower masses. This galaxy harassment can lead to stripping of the outer galactic halos, disk deformations in spirals and S0 galaxies, or trigger star formation activity (Richstone, 1976). 20 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

Ram pressure stripping: Objects moving through a gaseous medium with density ρgas 2 at velocity v are exerted to a ram pressure pram =ρgas vgal. A galaxy passing through the dense central region of the ICM can this way be eﬃciently stripped from its internal gas on time scales of 107–108 years. A single passage of a galaxy through a cluster core can hence almost completely remove the interstellar medium from this galaxy and quench its star formation.

One could expect that these additional physical eﬀects taking place in clusters modify or re-shape the galaxy luminosity functions (LF) of the cluster galaxy populations. However, the individual galaxy classes (e.g. E, S0, S) maintain a universal luminosity function with the same general Schechter function form as ﬁeld galaxies (e.g. Andreon 1998; Popesso et al., 2005)

L α L dL φ(L) dL = φ exp , (2.20) 0 L L L ∗ ∗ ∗ where L is the characteristic luminosity, φ0 the space density normalization, and α the faint-end∗ slope. This universal form of the luminosity function has been conﬁrmed out to redshifts of z 1.3 with typical slope parameters of α 1 0.3 for the bright part ≃ ≃ − ± of the composite LF and characteristic rest-frame absolute (AB) magnitudes of Ks 23.5 0.5 (Strazzullo et al., 2006). Note that the frequently used notation m*+1 refers∗ t ≃o the− apparent± (observed) magnitude of a galaxy with a luminosity that is one magnitude, i.e. a factor of 2.51, fainter than the characteristic luminosity L* at the given redshift. Similarly, M*-2 denotes galaxies which are two magnitudes brighter in absolute luminosity, i.e. a factor of 6.3, compared to a characteristic L* object of the given galaxy class.

2.4.2 Formation of ellipticals and the cluster red-sequence Early-type galaxies, i.e. elliptical (E) and lenticular (S0) galaxies, play a distinct role in galaxy clusters. Not only they constitute the dominant galaxy class in the core regions of clusters, but their properties exhibit a high degree of homogeneity, which allows to place strong constraints on their formation history. One of the key features of cluster galaxy populations is the existence of a tight sequence of red galaxies in the color-magnitude diagram (CMD), the so-called cluster red-sequence. This color-magnitude relation (CMR) in clusters has first been noticed by Baum (1959), but the full cosmological implications have only later been realized by Sandage & Visvanathan (1978). Bower, Lucey & Ellis (1992a; 1992b) confirmed the universality of the CMR in Virgo and Coma and established the red-sequence as distance indicator for clusters. The precision photometric study of these two clusters showed that the color-magnitude relations are intrinsically identical and that the observed color difference originates from the effects of the distance modulus (see Equ. 3.24). This study additionally established that the relation for the S0 population is less tight than for ellipticals, which exhibit a measured rms scatter along the red-sequence of 0.05 mag, mostly contributed by observational errors. The characteristic red-sequence with its minimal intrinsic scatter in low redshift clusters is 2.4. THE CLUSTER GALAXY POPULATION 21

of key importance for the understanding of galaxy evolution and can furthermore be used for an optical search and identification of new clusters (e.g. Gladders & Yee, 2005). The red-sequence in the CMD of the photometric bands X and Y can be characterized by (i) the color normalization (X Y) (Y*) at the location of L* galaxies, (ii) the slope − CMR of the sequence d(X Y)CMR/dY, and (iii) the (intrinsic) scatter σCMR. The color of the sequence is related to− the luminosity-weighted age of the stellar populations in galaxies and can be compared to galaxy evolution models for the derivation of a formation redshift estimate. The intrinsic scatter σCMR provides information on the star formation histories of the red-sequence galaxies as a tight relation should be only observable if these objects formed their stars over a period which is short compared to the age of the galaxies. The rising slope towards redder colors for more luminous objects can now be attributed to a metallicity sequence (Kodama and Arimoto, 1997). The required systematic increase towards higher metallicities and hence redder stellar populations10 can be quantitatively explained with supernova-driven wind models of ejected heated gas from the system. The deeper potential wells of massive galaxies, with increased gravitational binding energy per unit mass compared to smaller systems, can retain more metals and additionally trigger a more rapid metal production at early cosmic epochs. The explanation of the observed local CMR with stellar population evolutionary synthesis models was not straightforward due to a general age-metallicity degeneracy (Worthey, 1994). Galaxy colors become redder as a consequence of increasing metallicity or increasing age of the stellar population (in absence of dust). This degeneracy could be broken by moving to higher redshifts, where the two scenarios predict different redshift evolutions of the CMR. Models for an evolving metallicity sequence predict the observed (almost) redshift independent slope. An age sequence, on the other hand, would feature a rapid steepening and a CMR truncation at higher redshifts, since the formation epoch is ap- proached earlier for the faint younger blue galaxies than for the luminous older red ones (Kodama et al., 1998). The tightness of the red-sequence is expected to break down when the lookback time at the observed redshift z approaches the main epoch of star formation activity in (massive) early-type galaxies, i.e. when the stellar populations are ‘young’. However, all current known spectroscopically confirmed high-redshift galaxy clusters out to z 1.3 show a pronounced red-sequence with a surprisingly small scatter. Figure 2.5 displays∼ a recent Hubble Space Telescope (HST) color-magnitude diagram of the Lynx supercluster. Its two high-redshift clusters, the most distant ones discovered with ROSAT, exhibit intrinsic red-sequence scatters of merely 0.025 0.015 mag for the elliptical population (Mei et al., 2006b). From these observations, the± formation redshift of the red-sequence galaxies can be estimated as zf >2.5, yielding strong constraints for evolution models. Two competing formation paradigms for elliptical galaxies have been contrasted in the last few decades: the monolithic collapse model and the hierarchical formation scenario. The monolithic collapse model (Eggen et al., 1962) assumes that ellipticals and bulges

10This reddening with increasing metallicity is primarily caused by a lower temperature of the metal rich stars at the main sequence turnoﬀ (e.g. Renzini, 2006). 22 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

Figure 2.5: Color-magnitude diagram for the Lynx supercluster at z = 1.26, including the two most distant clusters known from the ROSAT era. Orange and red symbols indicate the individual clusters Lynx East and Lynx West. Elliptical galaxies are indicated by circles, S0s by squares, and spirals by triangles. Square boxes mark spectroscopically conﬁrmed cluster members, open circles show interlopers, i.e. foreground objects. Plot from Mei et al. (2006b).

formed at high redshift, when gas clouds collapse rapidly and turn instantaneously into stars. After this violent star formation burst, no further generations of stars are formed, implying that the stellar population properties of a galaxy follow a passive evolution, i.e. the luminosity changes only due to the aging of stars. Within the monolithic collapse model, elliptical galaxies have not changed their overall appearance, mass, or identity since the formation redshift zf , except for the ageing stellar population and the associated reddening of the spectral energy distribution (SED) with cosmic time. Note that essentially all observed cluster red-sequence colors are consistent with this simple galaxy formation model for bursts taking place at zf >3 (see Sect. 7.1.2). Conversely, the hierarchical formation∼ scenario (e.g. White & Reese, 1978; Kauffmann, 1996; Kauffmann & Charlot, 1998) rests upon the physically motivated standard structure formation and evolution paradigm discussed in Sect. 3.2. Hierarchical formation in this respect means that small galaxy-size dark matter halos form first and massive galaxies originate from subsequent merging of smaller units. Major merging events can significantly alter the overall galaxy properties and lead to morphological transformations, trigger star burst events, or start nuclear activity in the galaxy centers. The main difference to the monolithic collapse model, however, is the expected late appearance of massive spheroids, as they are assembled over cosmic time. Consequently, a rapid decline of the space density of the most massive systems with redshift is expected. Luminous ellipticals are relatively 2.4. THE CLUSTER GALAXY POPULATION 23

young objects in this scenario, mostly appearing at z <1, which has long been at odds with discoveries of massive spheroids at z>1.5 (e.g. Dunlop et al., 1996; Cimatti et al., 2004). An important aspect of hierarchical galaxy formation is the fact that the appearance and identity of objects can change over cosmic time due to merging and evolutionary eﬀects. This implies that the cluster elliptical population observed locally does not necessarily correspond to the ellipticals in high-redshift clusters, a consequence which has been named progenitor-bias (van Dokkum and Franx, 1996). A key for the consolidation of hierarchical models with observations is the strict distinction between the formation epoch of the stellar population in ellipticals and their mass assembly history (see Fig.2.7). If star formation after merger events is absent (dry merging), massive ellipticals can exhibit old, passively evolving stellar populations, while their mass and luminosity growth might still continue to the present. Recent progress in numerical simulations and semi-analytic modelling techniques have led to detailed predictions from the hierarchical formation scenario for the properties of the elliptical galaxy populations in clusters, now being largely consistent with observations (De Lucia et al., 2006). The star formation activity for the largest local model spheroids peaks at z 5, but only half of the systems possess progenitors with at least 50% of the ∼ ﬁnal mass at z 1.5, the lag between star formation epoch and assembly time increasing for more massive∼ ellipticals. Hence the leverage for an observational distinction between monolithic collapse and hierarchical formation scenarios is highest for clusters at z>1.

2.4.3 Evolution and environmental effects Environmental effects and galaxy evolution are closely interlinked since the discussed external physical mechanisms at work in galaxy clusters can drastically change the galaxy properties over cosmic time. One of the key features of the cluster environment is the existence of a pronounced morphology-density relation (Dressler, 1980), i.e. the observed correlation between the frequency of various Hubble types and the local projected galaxy number density. In the local Universe, the fraction of spiral galaxies decreases from 60–70% in field environments to about 10% in dense cluster cores. The elliptical population, on the other hand, increases its relative contribution from approximately 10% in the field to roughly 50% in cluster centers. The fraction of lenticular (S0) galaxies rises more moderately from 30% in low density environments to about 40% in high density ones. ∼ Significant changes of the morphological galaxy mix can already be observed at redshifts of z 0.5 (Dressler et al., 1997). From z = 0 to z = 0.5, the S0 galaxy fraction in cluster∼ centers has dropped by a factor of 2–3 to 15–20%, whereas the spiral fraction has proportionally increased to about 30–35%. Recent studies of the morphology-density relation at z 1 confirm this trend and measure a combined early type fraction of fE+S0 0.72 0.10,∼ which seems to be correlated with the bolometric X-ray luminosityh of thei≃ ± 0.33±0.09 cluster through the relation fE+S0 LX,bol (Postman et al., 2005). The rising fraction of ellipticals towards higher densities∝ remains essentially unchanged, implying that the population of elliptical galaxies has been essentially in place since z 1. Conversely, the ∼ decreasing S0 fraction with redshift suggests that an environment induced morphological 24 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

Figure 2.6: Variations of the properties of the elliptical galaxy population with cluster-centric distance as predicted at low redshift by the Millennium Simulation. Galaxy properties are extracted from massive dark matter halos with M > 8 1014 M . Going from the cluster center towards the tot × ⊙ outskirts, the age (lookback time) of the stellar∼ populations of ellipticals is predicted to decrease (top left), the metallicity to decrease (bottom left), and the typical galaxy colors to become bluer (top right). These trends are partly driven by mass segregation, i.e. the more massive galaxies are found close to the center (bottom right). From De Lucia et al. (2006). transformation from spirals to S0s occurs at redshifts of z < 1 and continues until the present epoch. ∼ This picture is qualitatively consistent with the Butcher-Oemler effect, which provided the first direct observational evidence for the evolution of the cluster galaxy population (Butcher and Oemler, 1978). The two-band photometry of clusters at z 0.4 revealed an excess of blue galaxies compared to local rich clusters. A more recent study∼ of 295 Abell clusters at z<0.38 confirmed the original findings and quantified the evolution of the blue galaxy fraction as f =(1.34 0.11) z 0.03, where objects more than 0.2 mag below the B ± · − red-sequence were defined as blue (Margoniner et al., 2001). Although the trend of such a relation is well established, the current form will obviously not hold at higher redshifts as it ‘predicts’ a vanishing of the red-sequence by z 0.8, at odds with observations (see ≃ Fig. 2.5). The galaxy colors are closely linked to their star formation rates (SFR), which can be probed as a function of galaxy density with the Sloan Digital Sky Survey (SDSS) or the 2dF survey for low redshift cluster environments (Lewis et al., 2002; Gómez et al., 2003). These studies showed the existence of a sharp transition in the density–SFR relation starting from field-like star formation rates at local projected galaxy densities of Σ 1 Mpc−2 and converging to the suppressed star formation activity of cluster cores at Σ >≃7 Mpc−2. The transition from field to central cluster SFR properties is hence completed∼ over a range of about seven in the local galaxy density and occurs roughly at the cluster virial radius. This large radius for the onset of star formation suppression, and the fact that a similar 2.4. THE CLUSTER GALAXY POPULATION 25

density–SFR relation is found for group environments, give rise to a scenario where the transition from star forming to passive cluster objects already occurs during the infall along filaments towards the cluster (Bower and Balogh, 2004). Using the high-resolution Millennium Run simulation (Springel et al., 2005), detailed predictions for the changes of galaxy properties as a function of cluster-centric distance can be obtained. The results for the elliptical cluster galaxy population at low redshift are displayed in Fig. 2.6 (De Lucia et al., 2006). Moving outwards from the high-density cluster centers, the simulations predict (i) decreasing ages, (ii) decreasing galaxy metallicities, (iii) bluer object colors, and (iv) decreasing stellar masses for the ellipticals. In other words, elliptical galaxies in cluster centers are expected to be older, more massive, redder, and more metal rich than their counterparts in lower density environments. These trends are partly driven by the dynamic friction induced mass segregation. Moving towards higher redshifts, a study of A 851 at z =0.41 yielded the first indication for a sharp radial transition in the properties of faint (<0.1 L*) galaxies (Kodama et al., 2001). Using deep photometric data, an abrupt change from a predominantly red galaxy population towards bluer galaxies was observed at projected densities of Σ 100 Mpc−2, corresponding to the location of the cluster subclumps in the filaments of the≃ surrounding large-scale structure (LSS). At medium distant redshifts, investigations of the LSS around a z =0.73 cluster in the COSMOS field (see Sect. 4.4) show a weak trend towards redder colors with increasing local density for early-type galaxies (Cassata et al., 2007). For RXJJ0910+5422 at z =1.11., Mei et al. (2006a) observe an S0 population that is systematically bluer than the elliptical red-sequence by about 0.1 mag, suggesting that this population is still evolving towards the redder CMR colors of the ellipticals. However, in the Lynx clusters of Fig. 2.5, ellipticals and S0 galaxies lie on the same red-sequence, but the scatter increases with distance from the center (Mei et al., 2006b). Recently Cucciati et al. (2006) and Cooper et al. (2007) have investigated the general color–density relation out to z 1.5 based on large spectroscopic galaxy surveys. These ∼ studies show that the fraction of red galaxies in dense environments gradually decreases with redshift, with a slower trend for the brightest objects. By redshifts of z 1.5, the segregation of red and blue galaxies in dense regions has completely disappeared.∼ These new results suggest that the maximal star formation activity shifts to galaxies with lower luminosities and in lower density environments with decreasing redshifts. This is possibly governed by a combination of biased galaxy formation in high density regions and environmentally induced star formation quenching. The observation that the star formation activity in the most massive galaxies ceases at earlier epochs is a manifestation of the so-called ‘downsizing’ scenario (Faber et al., 1995; Cowie et al., 1996). A direct consequence of ‘downsizing’ in high-redshift clusters would be the ‘truncation’ of the red-sequence at a given redshift dependent magnitude, i.e. an increasing deficiency of faint red galaxies with increasing lookback time. Galaxies move towards the red-sequence in the color-magnitude diagram after star formation activity has ceased. If this occurs at earlier epochs for the more luminous galaxies, then the red- sequence will be progressively populated over cosmic time starting from the bright end. Observations of the resulting deficit of faint red galaxies on the red-sequence have been claimed by De Lucia et al. (2004) for clusters at z 0.8 and by Tanaka et al. (2007) for a cluster at redshift z =1.24. ∼ 26 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

2.4.4 Brightest cluster galaxies

For distant cluster searches, the brightest cluster galaxies (BCGs) are of particular importance in the cluster identification procedure. In fact, (i) BCGs often mark the center of the associated galaxy overdensity, (ii) they usually have the best photometry and color information available among all cluster galaxies, (iii) the BCG location in the color-magnitude diagram usually marks the bright end of the cluster red-sequence and is hence important for the redshift estimate, and (iv) BCGs are prime targets for spectroscopic follow-up. The brightest cluster galaxies constitute a very special category of galaxies that still lacks a detailed understanding of the formation history. Compared to other galaxy classes, low redshift BCGs are more massive, have larger spatial extent, and are ultra-luminous with typical luminosities of LBCG 10 L (Schombert, 1986) that correlate with cluster mass in the form L M 0.26±0.04∼(Lin and∗ Mohr, 2004). In fact, the observed luminosity BCG ∝ 200 gap of the BCGs to the second ranked galaxies of massive clusters cannot be accounted for by a statistical sampling of the bright end of the Schechter LF (Equ. 2.20), indicating the need for a special formation process (Bhavsar and Barrow, 1985). In the local Universe, the first ranked cluster galaxies are predominantly located very close to the cluster’s DM potential well minimum; 75% are found within the central 3% of the virial radius. They are often classified as radio-loud objects and are associated with the cooling core phenomenon (Sect. 2.3.5) in a complex way. BCGs are also found to have a higher fraction of dark matter compared to similar massive objects resulting in larger velocity dispersions for the stellar component (von der Linden et al., 2007). More than half of all BCGs in local X-ray clusters are classified as cD galaxies, a galaxy type exclusively found in the centers of clusters. cD galaxies were originally defined by Matthews, Morgan & Schmidt (1964) as supergiant galaxies with an additional extended low surface brightness envelope in excess of the radial light distribution of a normal elliptical galaxy at large radii. The outer cD halo has no clear edge and can extend far into the cluster (Schombert, 1988). The most prominent formation scenario for these centrally dominating supergiant galaxies is the galactic cannibalism model of Ostriker & Tremaine (1975). In this picture, cD galaxies grow through cannibalism of their neighbors as the most massive galaxies sink to the cluster center driven by dynamic friction. The formation scenario for giant ellipticals forming via merging in cluster centers has been confirmed by numerical simulations (e.g. White, 1976), but the reproduction of the characteristic extended halo has remained challenging until the present (e.g. Nipoti et al., 2003). It seems likely, however, that the origin of the cD halo starlight is connected to tidally stripped material from the outer parts of galaxies or their tidal disruption (e.g. Zibetti et al., 2005). Despite their possibly ‘violent’ formation history, the brightest cluster galaxies constitute a surprisingly homogeneous class of objects with a low dispersion in their absolute magnitudes of MV BCG 24.5 0.3. Local BCGs have been characterized as good ‘standard candles’ exhibitingh i ≃− a Gaussian± scatter of the absolute magnitudes around the mean value (Postman and Lauer, 1995). Because of this characteristic, BCGs of massive clusters at redshifts z < 1 have been used for cosmological tests by applying the Hubble diagram of Sect. 3.3 (e.g.∼ Aragon-Salamanca et al., 1998; Collins & Mann, 1998). Over the redshift range covered, no signs of a significant BCG evolution have been found, other than the passive luminosity evolution (Burke et al., 2000). 2.4. THE CLUSTER GALAXY POPULATION 27

Figure 2.7: Mass assembly histories and formation times of BCGs as seen from the Millennium Simulation. Left: Stellar mass of the main BCG progenitor (lower black line) and the sum of all progenitors (top line) for a single case study BCG. The horizontal dashed line represents half of the ﬁnal BCG mass. Right: Mass assembly history (blue) and stellar formation epoch (green) for the full BCG sample. Plots from De Lucia & Blaizot (2007).

To complete this introductory section on the cluster galaxy population, the predictions of the state-of-the-art Millennium Run Simulation concerning the cosmic evolution of BCGs are presented, which have been recently discussed by De Lucia & Blaizot (2007). Figures 2.7–2.9 establish the simulated reference model for comparison with observations. The 125 traced model BCGs were selected from local cluster halos more massive than 14 7 10 M⊙ and have absolute K-band magnitudes of MK = 26.58 mag with a dispersion of× 0.2 mag. The galaxy assembly time is defined in theh analysisi − as the epoch when the single main BCG progenitor contains half of the final (z =0) stellar mass. At the formation time, on the other hand, half of the final total stellar mass is present in the sum of all progenitor objects. The predicted extreme hierarchical nature of the brightest cluster galaxies is illustrated in Fig. 2.7, which shows the mass evolution of a single case study BCG in the left panel and the averaged assembly history of the full sample in the right panel. The upper curves in both plots indicate that most of the BCG stars form at high redshift, 50% by z 5 and 80% ∼ by z 3, implying old, evolved, and red stellar populations as observed at the accessible redshifts.∼ The mass assembly history of the main BCG progenitor is displayed by the lower lines showing the discrete merging events for the single galaxy on the left and the average sample evolution on the right, with a predicted mean mass growth by a factor of three from z =1 to z = 0. Figure 2.8 provides testable predictions (see Sect. 11.1) for the evolution of the absolute rest-frame K-band luminosities (top panel) and after subtracting a passive luminosity evolution model with formation redshift zf = 5 (lower panel). The distribution of BCG stellar masses at different cosmic epochs is illustrated by the red histograms in Fig. 2.9. Observationally, the big questions when and how cD galaxies formed are yet to be explored. Between the spiderweb galaxy (Miley et al., 2006), a possible cD precursor in a protocluster environment at redshift z =2.2, and the well studied end products of the BCG evolution at z <0.8, there is still a lot of redshift space to fill in. ∼ 28 CHAPTER 2. GALAXY CLUSTERS AS ASTROPHYSICAL LABORATORIES

Figure 2.8: Top: Evolution of the absolute BCG K-band rest-frame magnitude as a function of redshift. Black dots represent model BCGs, red symbols indicate objects in halos more massive than 15 10 M⊙. The dashed lines show burst model predictions for a formation redshift of zf = 5 and ﬁxed mass. Bottom: Residuals after subtracting the burst model. The green shaded area indicates the redshift range which has been probed observationally. Plot from De Lucia & Blaizot (2007).

Figure 2.9: Evolution of the stellar mass distribution of high-redshift model BCGs (red histograms) and of progenitors of local BCGs (blue histograms). Plot from De Lucia & Blaizot (2007). Chapter 3

Galaxy Clusters as Cosmological Probes

Besides their use as astrophysical laboratories, galaxy clusters are also sensitive cosmological probes of increasing importance. The NASA Dark Energy Task Force report1, for example, recommends the study of cluster evolution as one of the four most promising pillars of future research in fundamental cosmology. The distant galaxy clusters under focus for this thesis are observed at an epoch which corresponds to the second quarter of the lifetime of the Universe. Their general observational properties have thus to be interpreted within the standard (homogeneous) cosmological model and modern structure formation scenarios. More complete and detailed general introductions to cosmology can be found in the textbooks by e.g. Schneider (2006a) and Coles & Lucchin (2002). This chapter pursues the following main motivations and aims: 1. Provide the general cosmological framework and tools for the consistent description of z>1 clusters within the concordance2 ΛCDM cosmology (Sect. 3.1); 2. Introduce the main aspects of modern hierarchical structure formation scenarios with the focus on cluster formation and evolution (Sect. 3.2); 3. Compile a summary of cosmological tests using (distant) galaxy clusters which are feasible within the next decade(s) (Sect. 3.3).

3.1 Cosmological Framework

3.1.1 Dynamics of the expanding Universe The dynamics of the scale factor R(t) of standard relativistic world models is governed by the Friedman–Lemâitre equations, which are derived from the field equations of general relativity (GR) for the case of a perfect homogeneous fluid

1http://www.science.doe.gov/hep/DETF-FinalRptJune30,2006.pdf 2ΛCDM means that the dominant energy components are cold dark matter (CDM) and Dark Energy.

29 30 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

¨ 4 3p 1 R = πGR ρ + 2 + ΛR (3.1) −3 c 3 8 1 R˙ 2 = πGρR2 + ΛR2 kc2 . (3.2) 3 3 − Here dots represent time derivatives of the scale factor R(t), ρ is the matter density, p the pressure, Λ Einstein’s cosmological constant, and k the space time curvature constant. The scale factor R(t) with the dimension of a length can be replaced by the dimensionless parameter a(t)= R(t)/R0, which is normalized to the scale factor of the current epoch R0, i.e. a(t0) 1. The two≡ equations for the expansion velocity R˙ (ora ˙) and acceleration rate R¨ (ora ¨) are not independent, but are linked3 through the adiabatic expansion equation (energy conservation in thermodynamics: dU = pdV ) − d d (a3ρc2)= p (a3) . (3.3) dt − dt From Equ. 3.3 the density evolution during the expansion of the Universe can be derived for the diﬀerent relevant energy components with their respective equations of state (EoS) p = wρc2. In ΛCDM cosmologies, the major energy components with their density scaling relations are:

matter (CDM and baryonic) : w =0 ρ a−3 (3.4) m ∝ radiation : w =1/3 ρ a−4 (3.5) r ∝ Dark Energy : w< 1/3 ρ a−3(1+w) (3.6) − DE ∝ The pressureless matter density (Equ. 3.4) is just diluted as space expands, the radiation component (and relativistic matter) experiences an additional dilution factor of a−1 due to the cosmological redshift (see Equ. 3.8) and thus its energy density decreases faster than matter. In the last step, the Dark Energy component was generalized to account for its unknown nature. For the special case of the cosmological constant Λ (w = 1), Equ. 3.6 0 − implies a constant energy density ρΛ a =const. The ratio of the expansion velocity∝ and the scale factor deﬁnes the Hubble parameter H(a), i.e. the fractional increase of the Universe per unit time a˙ H(a) . (3.7) ≡ a −1 −1 Its present value H0 H(a0) is the Hubble constant with H0 70 km s Mpc 2.3 10−18 s−1 (e.g. Freedman≡ et al., 2001). The lowest order approximation≈ of Equ. 3.7≈ yields× the famous Hubble law, which relates the distance D to the cosmological recession velocity

3To see the general relationship, the time derivative of Equ. 3.2 is taken and solved fora ¨. Reordering the right-hand side to match Equ. 3.1 leaves extra terms that can be identiﬁed with Equ. 3.3. 3.1. COSMOLOGICAL FRAMEWORK 31

vrecession (for small redshifts) via D =vrecession/H0. Hence, at constant velocity (or redshift) −1 all cosmological distances scale with the inverse of the value of the Hubble constant H0 , i.e. a faster expansion rate results in a smaller (comoving) distance at a given redshift. This distance scaling property can be parameterized with the dimensionless constant h = −1 −1 −1 −1 H0/(100kms Mpc ), or for a concordance cosmology as h70 = H0/(70 km s Mpc ). The dependency on the exact numerical value of the Hubble constant can then be accounted for by using the (parameterized) distance units [D]=h−1 Mpc. Analogously all secondary derived quantities that are linked to cosmological distances scale with the appropriate −3 3 4 −1 powers of h: volumes as [V ]= h Mpc , total cluster masses as [M]= h M⊙ , densities as [ρ]=h2 g cm−3, and luminosities as [L]=h−2 erg s−1. Since all distance scales are stretched by the expansion of the Universe, this also applies to the wavelength λ of light propagating through space. This leads to the cosmological redshift z, which is given by the ratio of the scale factors, i.e. the stretching factor, at the time of observation a0 and emission of the light a(t) λ a 1+ z obs = 0 . (3.8) ≡ λem a The redshift z is a direct observable and is hence of particular importance for observational cosmology. Through Equ. 3.8 redshifts are uniquely linked to the expansion state of the Universe and other quantities such as cosmic time and distances through the underlying cosmological model. Note that the cosmological redshift should not be interpreted as a recession velocity vrec as in special relativity theory (SRT), although for small redshifts (z < 0.3) the results are consistent as is the case for the Hubble law. At high redshift (z >∼1), however, it can be shown empirically (Davis and Lineweaver, 2004) that the SRT interpretation∼ fails and that super-luminal recession velocities are indeed allowed in the general relativistic treatment. In particular the Hubble radius dH, deﬁned as the sphere that presently recedes at the velocity of light

c −1 −1 dH = = 2997.9 h Mpc 4280 h70 Mpc , (3.9) H0 ≃ corresponds to a redshift of z =1.46 in concordance cosmology. This implies that objects at higher redshift recede super-luminally at the current epoch, and the furthest spectroscopically confirmed galaxies at z 6 have done so at all times5. As the next step, all energy densities∼ can be transformed into dimensionless parameters by stating them in units of the critical energy density ρcr. ρcr is defined as the total energy density that results in a flat global geometry of the Universe, i.e. k 0 in Equ. 3.2 (for Λ=0), and is thus given by ≡

3H2 ρ 0 =1.879 10−29 h2 g cm−3 =2.775 1011 h2M Mpc−3 . (3.10) cr ≡ 8πG × × ⊙ 4The distance scaling depends on the mass determination method and is referring to Equ2.13. The 5/2 X-ray determined cluster gas mass scales as Mgas h− (e.g. Sasaki, 1996). 5It is still possible to observe them, since the∝ Hubble sphere is receding even more rapidly and can ‘overtake’ the emitted photons, see Davis and Lineweaver (2004) for more details. 32 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

The dimensionless energy components as measured today are then deﬁned as

2 ρm ρΛ Λc ρr Ωm = ; ΩDE = = 2 ; Ωr = ; Ωk =1 Ωm ΩΛ Ωr . (3.11) ρcr ρcr 3H0 ρcr − − −

The total energy density is the sum of all contributing components

Ωtot = Ωm + ΩDE + Ωr . (3.12)

The first two terms for the matter density Ωm and the Dark Energy density ΩDE are the dynamically dominant contributions today. In concordance cosmology, their sum (see Tab. 3.1) is very close to the critical density Ω + Ω Ω 1., which implies a flat (or m DE ≈ tot ≈ very close to flat) Universe with Ωk 0 with percent level accuracy (Spergel et al., 2007). The radiation term Ω (10−4) is negligible≈ today but has been of significant importance r ∼O at early epochs of z > 1000 (see Sect. 3.2), which can be seen from the density scaling relations in Equ. 3.5. ∼ When combining equations 3.2, 3.7, and 3.11, the general Hubble expansion history is obtained as a function of the normalized scale factor

H2(a)= H2 Ω a−3 + Ω a−3(w+1) + Ω a−4 (Ω 1) a−2 . (3.13) 0 · m DE r − tot − h i Here the Dark Energy equation-of-state parameter w is left as a free parameter as in Equ. 3.6. However, the most general form of Dark Energy also allows for a time evolution of the equation-of-state w(z). In its simplest form the evolution can be parameterized as w(z)=w0 +w1 z with a constant part w0 and an evolving contribution w1. For an arbitrary time variation·w(z), the Dark Energy redshift scaling factor is integrated according to

′ z ′ z ′ z 1+w(z ) d (1+z′)3[w(z )+1] = exp d ln (1+z′)3[w(z )+1] = exp 3 dz′ . (3.14) 0 0 0 1+z′ ! Z h i Z h i Z For expressing all relevant relations in terms of the observable redshift z, the scale factors are substituted according to a (1 + z)−1 and da dz(1 + z)−2. As a final step, we → → − can now define the cosmic evolution function E(z), with which the final general form of the expansion history of the Universe H(z) can be expressed in terms of H0

′ z 1+w(z ) ′ 2 3 3 ′ dz 4 2 E (z) = Ω (1 + z) + Ω e 0 1+z + Ω (1 + z) + (1 Ω )(1 + z) (3.15) m DE · r − tot H2(Rz)= H2 E2(z) . (3.16) 0 · Detailed measurements of the Hubble expansion history H(z) are the core of many cosmological tests, in particular for future Dark Energy studies (see Sect. 3.3). 3.1. COSMOLOGICAL FRAMEWORK 33

3.1.2 Distance measures A meaningful general deﬁnition of distances in the Universe requires the knowledge of the metric for the global geometry. For the isotropic and homogeneous Universe (maximally symmetric space time) the general relativistic solution is the Robertson-Walker-Metric (RWM). The space time line element in spherical coordinates with curvature constant k and scale factor R(t) can be written as

dr2 ds2 = c2dt2 R2(t) + r2 sin2θdφ2 + dθ2 (3.17) − "1 kr2 # − with proper time t, radial comoving coordinate r, and angular coordinates φ and θ. As the global space time metric, the RWM is valid for all distances. For many applications the light propagation along radial coordinates is of prime interest. Since light travels on null geodesics, i.e. ds2 =0, Equ. 3.17 reduces to the simple form

dr2 c2dt2 R2(t) =0 . (3.18) − 1 kr2 − For the following discussion of distance measures, the specialization for a flat Universe (k 0) is assumed, which is valid for concordance cosmology. Small deviations from a flat ≡ geometry could still be possible, however, the additional correction functions sin(R0r) for k = 1 and sinh(R r) for k = 1 for non-flat space times would inhibit the readability and 0 − clarity for this discussion and are therefore omitted. Integrating Equ. 3.18 (for k = 0) yields the fundamental comoving distance D, i.e. the distance an object at redshift z has today,

r t0 ′ z ′ z ′ ′ dt dz c dz D(z)= R0 r(z)= R0 dr = c ′ = c ′ = ′ . (3.19) · Z0 Zt a(t ) Z0 H(z ) H0 Z0 E(z ) The second last step took advantage of Equ. 3.8 and its time derivative dt = dz(1+ − z)−1H(z)−1 to relate time t and redshift z, and the final step made use of Equ. 3.16. From the observational point of view, two other alternative cosmological distance measures are of prime importance. Firstly, the luminosity distance dlum relates the luminosity L of an object to its observed flux f in a way to reproduce the inverse square law of flat Euclidean 2 space f =L/(4πdlum). Secondly, the angular diameter distance dang is defined to yield the Euclidean relation dl = θ dang for the apparent angular size θ of an object with physical size dl. Both observationally· motivated distance measures can be expressed in terms of the comoving distance D as

d = D (1 + z) (3.20) lum · D d = . (3.21) ang (1 + z) 34 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

The three main cosmological distance measures D, dlum, and dang are illustrated in the left panel of Fig. 3.16 for concordance model parameters. The fact that the angular size distance dang has a maximum at z 1.6 has the important observational consequence that the apparent angular size of objects≈ is practically constant at z>1, as shown in the right panel of Fig. 3.1. This implies that high-redshift galaxies with typical physical sizes of 10 kpc are observed at angular scales of 1.2′′, 100 kpc cluster cores at 12′′, and 1 Mpc scale clusters of galaxies at 2′. 2 Combining the relation dlum = dang (1+z) and the fact that the solid angle dΩ covered 2 · by an object follows dΩ dang leads to the cosmological surface brightness dimming (Tolman’s Law) ∝

I (emitted) I (observed) = bol . (3.22) bol (1 + z)4 For monochromatic flux measurements, the surface brightness dimming reduces to I (observed) (1 + z)−3, since the factor (1 + z)−1 for the effective narrowing of the mono ∝ observed rest-frame bandwidth is recovered. If the rest-frame spectral energy distribution (SED) has a non-zero slope at the spectral interval of interest, then the observed flux will not only depend on the luminosity distance dlum but also on the actually observed part of the rest-frame spectrum redshifted by the factor (1 + z). This is taken into account with the so-called K-correction, which can be defined as an additive term KQR for the observed apparent magnitude mR in band-pass R

mR = MQ + DM + KQR , (3.23) where MQ is the absolute magnitude of the source in the rest-frame band-pass Q and DM is the distance modulus

DM = m M = 25+5log(d [Mpc]) 5log h . (3.24) − lum − 70 This last relation follows directly from the definition of absolute magnitudes transformed −1 7 to h70 Mpc units . For concordance cosmology, the last term is zero, and dlum can be read off from Fig.3.1. The K-correction term KQR depends on the filter band, the redshift, and the spectral properties of the object under investigation (see Fig. 11.4). Positive correction terms imply a resulting K-dimming due to a decreasing SED with increasing frequency; negative terms result in K-brightening, i.e. objects actually appear brighter than expected at higher redshift. For efficient observations of the high-redshift Universe, the K-correction is an important consideration for the observing strategy as discussed in Sect. 7.1. With the distance measures at hand, it is now possible to define comoving volume elements dVcom as

6The models for Figs. 3.1, 3.2, and 3.3 were computed with the cosmological calculators at http://www. astro.ucla.edu/%7Ewright/CosmoCalc.html and http://faraday.uwyo.edu/%7Echip/misc/Cosmo2/ cosmo.cgi. 7The last term vanishes for concordance cosmology, but reﬂects the constant oﬀset when changing the value of the Hubble constant H0 3.1. COSMOLOGICAL FRAMEWORK 35

cD2 dV = D2dΩ dr = dΩ dz . (3.25) com H(z) Here D is the comoving distance, dΩ the solid angle, and dr the radial thickness8 of the volume element. In the last step, the radial component was expressed in terms of a redshift interval dz as in Equ. 3.19. Figure 3.2 (right panel) displays the enclosed comoving volume out to redshift z per square degree for concordance cosmology parameters. The left panel depicts the observed (bolometric) flux for a selection of fiducial object luminosities as a function of redshift and illustrates the luminosity dependent search volumes for flux-limited observations. When integrating a homogeneously distributed population of sources in Euclidean space over all luminosities (with arbitrary luminosity function), one obtains a universal relation for the cumulative number counts N(>S) above flux S9

− 3 N(>S) S 2 . (3.26) ∝ The log N(>S)–log S plot thus shows a constant slope of 3/2 down to the flux limit of − the observation, where the counting becomes incomplete (see also Equ. 3.45). The slope is due to the fact, that a lower flux reaches out further in distance (for any luminosity − 1 interval) d S 2 , and the enclosed volume grows proportional to the third power of lum ∝ the distance. Since this is only valid for static flat Euclidean space, the relation is expected to intrinsically flatten at a certain flux level when the effects of the expanding Universe and evolving populations become significant (see Sects 3.3 & 4.3). For a population of objects with a constant comoving number density n0 = dN/dVcom, the redshift distribution dN/dz is obtained by using Equ. 3.25

dN dV c D2 = n com = n dΩ . (3.27) dz 0 dz 0 H(z) The unique link between redshift z and cosmic distance measures can also be generalized for cosmic time t. The time derivative of Equ. 3.8 yieldsa ˙ = (1+z)−2 dz/dt. Dividing − by a(t) and substituting with relation 3.7 results in H(z)= (1+z)−1 dz/dt, which can be − readily integrated by separation of variables to arrive at the lookback time tlookback

t z ′ ′ 1 dz tlookback(z)= dt = ′ ′ , (3.28) t0 H 0 (1 + z ) E(z ) Z 0 Z · i.e. the time light from an object at redshift z has been travelling (light travelling time). Integrating to z = returns the age of the Universe t , and the diﬀerence yields the age ∞ 0 of the Universe at a given redshift t = t t (z). Fig. 3.3 illustrates t and t age 0 − lookback age lookback for a concordance cosmology with t0 =13.46 Gyr. Table 3.1 summarizes the main cosmological parameters of the concordance model following Spergel et al. (2007). These parameters will be used throughout this thesis for any background cosmological model unless otherwise stated.

8 Strictly speaking d(R0r), as the radial coordinate r is dimensionless. 9Same meaning as ﬂux f, but S is typically used in the context of the log N–log S relation. 36 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

Figure 3.1: Cosmological distances and scales for objects at redshift z using concordance model parameters. Left: Comoving distance D (blue solid line), luminosity distance dlum (black), and angular size distance dang (red) as a function of redshift. The Hubble radius dH is indicated by the horizontal dashed line and corresponds to the comoving distance at redshift z =1.46. Right: Physical and angular scale evolution with redshift. The black line depicts the physical comoving size of a ﬁxed angular scale of 1′′ in units [kpc/arcsec], the blue line traces the apparent angular size of a ﬁxed physical scale of 1Mpc in units [arcmin/Mpc]. Due to the global maximum of d at z 1.6, both functions ang ≈ assume their extrema at this redshift. However, at z>1 the scales are almost redshift independent at approximately 8.4 kpc arcsec−1 and correspondingly 2.0 arcmin Mpc−1.

Figure 3.2: Bolometric flux evolution and comoving volumes as a function of redshift. Left: Received bolometric flux for three fiducial objects at redshift z with luminosities 1043 erg s−1 (red line), 1044 erg s−1 (blue line), and 1045 erg s−1 (black line). A limited detection bandwidth will shift the curves down with an offset that depends on the flux fraction outside the filter band. At a given flux limit (e.g. 10−14 erg s−1cm−2 for the dashed horizontal line) objects are detectable out to a maximum redshift zmax, which is increasing with luminosity. Right: Comoving volume Vcom(< z) per square degree sky coverage from redshift 0 to z. The search volume increases roughly by a factor of 5 from redshift 0.5 to 1.0, and doubles again out to z =1.5. 3.1. COSMOLOGICAL FRAMEWORK 37

Figure 3.3: Cosmic time as a function of redshift for concordance model parameters. Left: The black solid line indicates the age of the Universe tage at redshift z with the total age t0 = 13.46 Gyr as represented by the horizontal dashed line. The blue line illustrates the lookback-time tlookback from the current epoch to redshift z. The second quarter of the lifetime of the Universe (6.8Gyr < ∼ tlookback < 10.1Gyr) is spanned by the redshift range 0.8 < z < 1.9; from redshift z = 1 to z = 1.5 ∼ ∼ ∼ the lookback-time increases from 7.7 Gyr to 9.3 Gyr. Right: Redshift z versus lookback time tlookback (black line with left scale). Many physical processes related to structure formation have characteristic timescales of 1 Gyr, which translate into redshift in a non-linear way. The blue line (right scale) ∼ illustrates the changing redshift interval per Giga year lookback time dz/dt. Out to z <0.6 the redshift intervals are almost constant at 0.1 Gyr−1, i.e. redshift and time are linear to ﬁrst∼ order. By z = 1 ∼ the redshift interval per Giga year has increased to 0.25 Gyr−1, at z = 1.5 to 0.4 Gyr−1, and at z = 2 it is 0.6 Gyr−1.

Parameters Value Parameters Value

matter density Ωm =0.3 total energy density Ω0 =1 power spectrum normalization σ8 =0.8 primordial PS slope nS =1 Dark Energy density ΩΛ =0.7 age of Universe [Gyr] t0 = 13.46 Dark Energy EoS w = 1 recombination redshift z 1100 − rec ∼ Hubble constant [km s−1 Mpc−1 ] H = 70 matter-radiation equality z 3500 0 equ ∼ baryon density Ωb =0.045

Table 3.1: Cosmological concordance model parameters as used throughout this thesis. Left: Pa- rameters for which galaxy cluster samples are particularly sensitive to. Right: Additional concordance model parameters that will be assumed for the background cosmology. The parameters σ8 and nS describe the properties of the initial power spectrum of density perturbations, all other parameters are linked to the global geometry of the Universe, which is assumed to be ﬂat (k =0) in concordance cosmology. The chosen parameter values follow Spergel et al. (2007). 38 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

3.2 Cosmic Structure Formation

Whereas Sect. 3.1 set the framework for the homogeneous Universe, this part will introduce some relevant aspects of the inhomogeneous Universe, i.e. the formation and evolution of cosmic structure. In the hierarchical structure formation scenario, galaxy clusters play a distinct role as the largest and most recently virialized objects to form and mark the important transition regime between linear and non-linear gravitational dynamics. The standard paradigm for the growth of density perturbations rests upon the following qualitative picture: (i) Quantum fluctuations in the very early Universe serve as initial density fluctuations. (ii) These fluctuations are stretched to macroscopic scales by an early inflationary phase which also imprints a characteristic primordial fluctuation power spectrum. (iii) After inflation, density perturbations grow only through gravitational instability within the expanding background cosmology. From this point on, the structure evolution can be well traced with large numerical simulations using cold dark matter (CDM) as the main ingredient. (iv) The structure evolution in the non-linear regime and the increasing importance of complex non-gravitational physics (‘ga(s-a)strophysics’) on small scales is only accessible with elaborate numerical experiments. This regime includes all relevant galaxy scales and hence turns the detailed understanding of galaxy formation and evolution into a very difficult and unsolved task. However, down to the galaxy cluster scale of 10 Mpc analytic approximations and models can still provide the basic structure ∼ formation framework and valuable physical insights. In preparation for the cosmological tests presented in the next section and to provide a basic understanding of the formation process of galaxy clusters, we will summarize four main aspects of structure formation theory with a focus on galaxy clusters: (i) the general growth of density perturbations, (ii) the top-hat collapse model, (iii) the Press-Schechter mass function, and (iv) the spatial distribution of clusters.

3.2.1 Growth and collapse of density perturbations Perturbation Growth We will first have a closer look at how the primordial density fluctuations seeded by the inflationary epoch grow in general. This growth is mainly driven by the cold dark matter density component ρCDM which only interacts gravitationally. In the early Universe, however, the strong coupling of photons and baryons (the photon-baryon fluid) has important consequences on the structure formation. The homogeneous Dark Energy density ρDE, as the fourth player of perturbation growth, has an additional indirect influence by governing the dynamics of the expanding background cosmology. The hydrodynamic equations are the starting point for the description of the gravitational instability of density fluctuations. Besides Euler’s Equation (momentum conservation), the continuity equation (mass conservation) and Poissions’s Equation are needed to close the system. These equations are transformed to comoving coordinates and the matter density field ρ(x, t)=ρ ¯ (t) [1 + δ(x, t)] is expressed through the dimensionless m · 3.2. COSMIC STRUCTURE FORMATION 39

density contrast δ(x, t)

ρ(x, t) ρ¯ (t) δ(x, t)= − m . (3.29) ρ¯m(t) The system can then be linearized for small fluctuation amplitudes δ 1 on top of the | ≪ | mean matter density fieldρ ¯m and separated into the homogeneous background solution for the Hubble expansion (Equ. 3.4) and the inhomogeneous solutions for the evolution of the density contrast δ. The result of the perturbation analysis is a second order differential equation that governs the gravitational amplification of the fractional density contrast δ, which can be expressed in its general form as a damped wave equation

2 2 a˙ cs k δ¨ +2 δ˙ = δ 4πGρ¯m . (3.30) a " − a2 # The left hand side contains the acceleration and the damping term, where the Hubble expansion H =a/a ˙ acts as a friction force or a Hubble drag. The right hand side of Equ. 3.30 shows the gravitational force due to density fluctuation and an additional term originating from Euler’s Equation when considering fluids with pressure, as is the case for the photon-baryon fluid of the early Universe. The speed of sound for this relativistic 2 2 fluid with equation-of-state according to Equ. 3.5 is then given by cs = ∂p/∂ρ c /3. For sufficiently large perturbation wave numbers k =2π/L, corresponding to small≃ wavelength scales L, the right hand side is negative implying an oscillating solution. At this point, we will focus on the growing solution, which is obtained for the large scale perturbations with k (a/cs)√4πGρ¯m in the early Universe or as the general perturbation growth mode after matter-radiation≪ decoupling, when the baryonic pressure vanishes. For a negligible oscillation term on the right hand side, Equ. 3.30 does not contain any spatial derivatives implying that the time evolution for the perturbation growth is separable from the initial spatial perturbation field δi+(x, ti) according to

δ(x, t)= D (t) δ (x, t )+ D (t) δ (x, t ) . (3.31) + · i+ i − · i− i The decaying solution D−(t) is of little physical interest and will not be considered. The general expression for the growing perturbation solution D+(t) (e.g. Borgani, 2006 ) in linear approximation as a function of redshift z is given by

∞ ′ 5 1+ z ′ D+(z)= ΩmE(z) ′ 3 dz . (3.32) 2 · Zz E(z )

In an Einstein-de Sitter Universe (Ωm = 1) the perturbation amplitudes grow proportional −1 2/3 to the scale factor D+ a (1+z) (t/ti) . For an empty Universe (Ωtot 0), (virtual) perturbations are frozen∝ and∝ do not∝ grow at all. In a low density Universe (≡e.g. Ω 0.3, m ∼ ΩDE = 0) the driving gravitational force on the right hand side of Equ. 3.30 is decreased resulting in a slower growth rate at low redshifts. At early epochs, however, the matter density of such models is still close to critical, i.e. to the Einstein-de Sitter case, with an analogous fast early growth phase which slows once the matter density drops signiﬁcantly 40 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

below the critical value. When looking backwards in time from the present perspective, the eﬀect of this slowed structure evolution is an increased number of objects, e.g. clusters, at high redshift relative to a critical matter density Universe. For a concordance cosmological model (Ωm 0.3, ΩDE = 0.7), the addition of a Dark Energy component results in an intermediate∼ degree of evolution, since the Hubble expansion must have been slower in the past, relative to an open Universe, leading to a smaller friction term in Equ. 3.30. Note that the ‘Hubble drag’ term depends on the cosmological parameters, which implicitly inﬂuence the perturbation growth in addition to the explicit dependance on the mean matter density ρ¯m.

Top-hat collapse An important approach to tackle the non-linear structure evolution regime is provided by Birkhoff’s theorem, which states that a closed sphere within a homogeneous Universe evolves independent of its surroundings, i.e. as if no external forces are exerted on the sphere10. This implies that any overdense region of the Universe can be conceived as a homogeneous mini-Universe with the evolution driven by the local density parameters of the region under consideration. The spherical top-hat model is the only system for which the collapse of an overdense region of the Universe can be treated analytically. For this we consider a spherical region with radius R and a homogeneous overdensity δTH embedded in a background field with constant densityρ ¯m. According to Birkhoff’s theorem the evolution of this region is independent of the background field and can be described by the Newtonian approximation of the first Friedmann equation 3.1, i.e. with vanishing pressure and zero cosmological constant term. In analogy to a closed Universe, this finite region will expand up to a maximum radius Rmax at turn-around time tturn and then re-collapse in a time-symmetric fashion at time t =2 t . In practice, the overdense region will not collapse to a point but stabilize col · turn in a bound dynamic equilibrium state at radius Rvir Rmax/2 once the Virial condition ≃ 11 2 E¯kin+E¯pot =0 is fulfilled. The gravitationally bound, virialized object with total energy · 2 Etot = E¯pot/2= 3GM /(5 2Rvir) is now detached from the Hubble flow, i.e. it does not take part in the− overall expansion· of the background cosmology. In the case of an Einstein-de Sitter cosmology with Ωm =1 some universal characteristics of the collapse process can be specified, with only small variations for other cosmologies. The density contrast at turn-around time tturn, i.e. at the time of decoupling from the Hubble flow, is about δ 4.55. At collapse time t the contrast has increased to its final turn ≃ col (universal) equilibrium value12 of δ 177, which gives rise to the widely used definition vir ≃ of R200 as the cluster radius with an average density of 200 times the critical density ρcr. In comparison, the linear extrapolation of the density contrast according to Equ. 3.32 at t yields δ (linear) 1.69, and at turn-around δ (linear) 1.06, emphasizing the onset col vir ≃ turn ≃ 10In reality residual tidal forces are expected which can be neglected for this discussion. 11A factor of 3/5 arises for the potential energy of a sphere with uniform density. 12This density contrast is roughly obtained from the turn-around density by considering the compression factor during collapse and the evolution of the critical density during the formation process. 3.2. COSMIC STRUCTURE FORMATION 41

of the non-linear regime and the break-down of the linear approximation during object −1 collapse. Note that a virialized cluster with final radius Rvir 1.5 h70 Mpc originates from −1 ≃ a collapsed region of size 6 Rvir 10 h70 Mpc. These considerations imply∼ · that≃ any spherical overdensities reaching the critical density contrast δ detach from the Hubble flow and form a virialized object at t =t 2 t . turn vir col ≃ · turn For the cold dark matter standard model with a scale invariant initial perturbation spectrum, sub-galactic halos are the first to decouple from the Hubble flow, since perturbations on small length scales have greater amplitudes (see Equ. 3.36). Galaxy clusters on the other hand have formed relatively recently and hence appear late on the cosmic stage, marking the largest mass scales that have had sufficient time to virialize. The top-hat collapse model enables important physical insights into the cluster formation process. However, it rests upon many simplifications (spherical symmetry, homogeneous mass distribution, no external tidal forces, only gravitational physics) which are at best only an approximate description of the cluster formation conditions. The Millennium Run simulation of Springel et al. (2005) provides the most detailed and realistic cluster formation scenarios currently available. Figure 3.4 shows simulation snap-shots of the formation of a rich galaxy cluster over the last 10 Giga years of cosmic time (2 z 0). Besides the detailed dark matter distribution in the left panels, these simulations≥ also≥ trace the intracluster medium gas density (center panels) and the ICM temperature (right panels), which are directly accessible through X-ray observations. Note that the ICM gas density is a very good tracer of the underlying dark matter distribution, and that the substructure and asymmetry increases drastically beyond z > 1. ∼ 3.2.2 Galaxy cluster statistics After discussing the principal formation process of individual objects of the large-scale structure, we will now turn to the statistical properties of the overall galaxy cluster population. This section will focus on two essential properties of clusters which provide a sensitive link to cosmological models: (i) the distribution of cluster masses (mass function), and (ii) the spatial clustering properties (power spectrum).

Cluster mass function The basic framework to quantitatively approximate the comoving number density of (relaxed) dark matter halos as a function of mass and redshift n(M, z) is provided by the Press-Schechter formalism (Press and Schechter, 1974). The main idea of this procedure is the statistical evaluation of the Gaussian fluctuation field δ(x, t) and the identification of density peaks above the critical collapse value after applying a spatial filter function corresponding to a minimum object mass Mmin.

13The simulation movie, from which the snap-shot images are taken, can be found at http://www. mpa-garching.mpg.de/galform/data vis/index.shtml. 42 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

Figure 3.4: High resolution simulation of the formation of a rich cluster of galaxies from the Millennium Run. From top to bottom snapshots at redshifts of 2.0, 1.5, 1.0, 0.5, and z =0 are shown. The left panels illustrate the dark matter distribution, the center panels the gas density, and the right panels the temperature structure at the corresponding redshift. Note the signiﬁcant increase of substructure and asymmetry with redshift. The gas density and temperature are in principle accessible cluster observables by means of X-ray observations. Credits: V. Springel.13 3.2. COSMIC STRUCTURE FORMATION 43

Since the filtering concept will be of central importance for this section, we will first have a closer look at the properties of smoothed density fields. Let us denote WR(x) as a (3-D) spatial window function with smoothing length scale R. In the simplest case of a top-hat filter this will just be a sphere with radius R, which has a zero value outside and a normalized constant value W ( x R)=(4πR3/3)−1 inside the sphere. A filtered density R | |≤ field will hence contain the average value within this sphere surrounding any given point x. Another widely used window function is a Gaussian filter, but more complicated functional forms are conceivable. Mathematically spatial filtering is equivalent to a convolution of the density field δ(x) with the window function WR

δR(x)= δM(x)= δ(y) WR( x y ) dy , (3.33) Z | − | where δR(x) denotes the filtered field. In the case of the spherical top-hat filter, the average mass M contained inside the window function is M 4πR3ρ¯ /3, providing a one-to-one h i≃ m correspondence between the comoving smoothing length scale R and the mass scale M. Since a spatial convolution translates into a simple multiplication in k-space, working with the Fourier representation of the density contrast is often advantageous

˜ 1 ikx δ(k)= 3/2 dx δ(x) e . (3.34) (2π) Z

The filtered density contrast in Fourier space is then δ˜R(k) δ˜(k) W˜ R(k), with W˜ R(k) being the Fourier transform of the window function. The averaged∝ squared· density fluctuations in Fourier space define the important power spectrum P (k), or more accurately, the power spectral density function

˜ 2 1 2 sin(kr) P (k) δ(k) = 2 drr ξ(r) . (3.35) ≡ h| | i 2π Z kr The right hand side shows that the power spectrum P (k) is in effect the Fourier transform of the 2-point correlation function ξ(r), which states the probability to find two objects at separation r in excess to a random distribution. The power spectrum expresses the amount of cosmic structure as function of wavenumber k corresponding to the length scale L 2π/k. As a second order statistic, the power spectrum provides a complete statistical characterization≃ of the underlying Gaussian14 fluctuation field, in analogy to the mean and standard deviation as parameters for the one dimensional Gaussian distribution. Combining the discussions on spatial filters and Fourier representations, we can introduce the variance of the smoothed fluctuation field δR(x) as

2 2 2 ˜2 1 2 ˜ 2 σR = σM = δR = δR = 2 dkk P (k) WR(k) . (3.36) h i h i 2π Z Here Parseval’s theorem15 was applied and in the last step, Equ. 3.35 and the isotropy assumption were used, implying that the ﬁnal result can only depend on the norm of

14Gaussianity is one of the underlying assumptions for the Press-Schechter theory. 15The norm of a function is invariant under Fourier transformation, i.e. f(x) 2dx = f˜(k) 2dk . | | | | R R 44 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

the wave vector k but not its direction. As discussed, the smoothing scale R is directly related to the mass scale M and thus to the mass variance σ2 δM/M 2 , with the exact M ≡h| | i functional form depending on the shape of the used filter. We can now return to the Press-Schechter theory and outline the procedure to obtain an approximate dark matter halo mass function n(M, z), which can be interpreted as galaxy cluster mass function. In addition to the spatial dependence of the fluctuation field, the time evolution, or equivalently the redshift dependence, has to be considered according to the linear theory evolution of Equs. 3.31 & 3.32. As pointed out, the collapsed objects reach a virialized state once a critical linearly extrapolated density contrast δc =δvir(linear) 1.69 is reached, with a weak dependance of the exact numeric value on the cosmological≃ model and the redshift. For a linear extrapolation to the present time t0 or z = 0, density contrast peaks with amplitude δc can be identified with objects currently reaching the virial equilibrium state, i.e. presently formed systems. On the other hand, objects that have already collapsed and formed at higher redshift zform will have a higher density contrast today δ0,c =δc/D+(zform) (see Equ 3.31). The linear growth factor D+(z) with normalization D+(z =0) 1 is hence the main factor for the redshift evolution of collapsed objects. For an Einstein-de≡ Sitter Universe the last expression reduces to δ (EdS)=1.69 (1 + z ), but 0,c · form generally both the critical overdensity δc and the linear growth factor D+(z) are cosmology dependent. In order to obtain the number of collapsed objects at redshift z with mass M the following procedure is applied: (i) Determine the smoothing length scale R corresponding to the mass scale M. (ii) Filter the present fluctuation field with the window function WR to arrive at δR(x, t0). (iii) Count the smoothed density peaks for amplitudes above δc at redshift z, i.e. for the minimum present amplitude δmin >δ0,c =δc/D+(z). These peaks are interpreted as collapsed objects of mass M at redshift z. The probability p(δ ) to find ≥ 0,c amplitudes above the collapse threshold δ0,c follows from Gaussian statistics and is given by

∞ 1 δ0,c p(δ0,c, R)= p(δ, R) dδ = erfc , (3.37) δ0,c 2 √2 σ ! Z · R 16 where erfc(x) is the error function and σR = σM the smoothed fluctuation amplitude of Equ. 3.36. This probability can be interpreted as spatial filling factor or mass fraction of collapsed objects above the threshold mass. The final Press-Schechter mass function n(M, z)(e.g. Borgani, 2006 ), i.e. the redshift evolution of the dark matter halo number density, is derived from the above filling factor of collapsed objects at redshift z in the mass interval [M, M +dM] as

dn (M, z) ∂p 2 2 ρ¯ δ d log σ (z) δ2 PS = = m c M exp c . (3.38) dM −∂M V sπ · M 2 · σ (z) · d log M · −2 σ2 (z)! M M M

2 2 16 ∞ t The error function is deﬁned as erfc(x)= √x x e− dt. R 3.2. COSMIC STRUCTURE FORMATION 45

To go from a filling factor per unit volume to a cluster number, one has to divide by the occupied volume of the object V M/ρ¯ . The additional factor of two corrects for a M ≃ m shortcoming of the Press-Schechter theory namely that the integral of Equ. 3.37 over all mass scales only accounts for half the total mass (the ‘cloud-in-cloud’ problem). The mass dependence of the probability p(δ0,c) is implicitly contained in the mass-scale smoothed amplitude σM of the fluctuation field. This important factor also contains the dependence on cosmological parameters through the linear growth factor D+(z) governing the σM evolution with redshift. This implies that the high-mass end of the cluster mass function, where the last term dominates, is ‘exponentially sensitive’ to variations of cosmological parameters. The Press-Schechter theory in the form presented here rests upon several strong assumptions, e.g. a spherically symmetric collapse without substructure, and some ad hoc procedures such as the correct filter form or the cloud-in-cloud problem. Meanwhile extended Press-Schechter formalisms provide more rigorous derivations of the mass function (e.g. Bond et al., 2001; Bower, 1991) . Sheth & Tormen (2001) have generalized the mass function for the more realistic assumption of an ellipsoidal collapse. A particularly well 12 15 tested and widely used fitting formula for the mass range 10 –10 M⊙ has been given by Jenkins et al. (2001) based on large numerical simulations. A comparison to the more precise Jenkins mass function

dn (M, z) ρ¯ d ln σ−1 J =0.315 m M exp ln σ−1 +0.61 3.8 (3.39) dM · M 2 · d ln M · −| M | reveals that the Press-Schechter mass function underestimates the number density of massive clusters and overestimates the abundance of low-mass objects. Figure 3.5 illustrates the results of the Millennium Run simulation and the very good agreement with the predictions of the Jenkins mass function. Shown is dark matter halo 2 multiplicity function (M /ρ¯m) dn/dM = N (M/dM) (VM /V ), i.e. the mass fraction carried by objects in a unit mass bin. Note that the· density· from z =0 to z =1.5 drops by three orders of magnitude for halo masses of 1015 h−1M and by one order of magnitude for ∼ ⊙ masses of 1014 h−1M . ∼ ⊙

Cluster power spectrum As a last application of the standard structure formation scenario, we will take a closer look at the spatial distribution of galaxy clusters, i.e. the clustering properties of clusters. Figure 3.6 shows the simulated cosmic web of the large-scale structure on a 500 Mpc scale (left) and zoomed in by a factor of four (right). The top panels show the simulated dark matter distribution at redshift z =1.4 and the bottom panels the present day structure. The main aim of this section is to introduce the basic tools for a quantitative statistical description of the large-scale structure and how the observed spatial galaxy cluster distribution is connected to cosmological parameters. The tools and tests of this subsection will

16http://www.mpa-garching.mpg.de/galform/virgo/millenium/index.shtml 46 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

Figure 3.5: Differential dark matter halo number density as a function of mass and epoch as ob- 2 tained with the Millennium Run simulation. Shown is the halo multiplicity function (M /ρ¯m) dn/dM against the halo mass for different redshifts. Of particular interest are the observationally accessible curves at redshift z = 0 and z = 1.5. Solid lines show predictions using the Jenkins fitting formula (Equ. 3.39), dotted lines show Press-Schechter model predictions of Equ. 3.38. Plot from Springel et al. (2005).

not have direct applications for the current XDCP survey but will be of prime importance for the next generation of galaxy cluster surveys (see Chap. 12), for which XDCP assumes a pathfinder role for high-redshift clusters. The power spectrum P (k) is the main tool for the statistical description of the large- scale structure and has been introduced in Equ. 3.35. Whereas the overall shape and features of P (k) can be theoretically derived, the amplitude or normalization of the power spectrum has to be empirically determined. This can be achieved with the help of σ8, a very important but probably the least intuitive cosmological parameter. For a given filtering length scale R, Equ. 3.36 provides the link between the mass vari- 2 ance σR on this scale and the power spectrum P (k), which fixes the a priori unknown absolute normalization of P (k). The chosen filter scale R should on one hand be sufficiently large to ensure the validity of linear theory and on the other hand small enough 3.2. COSMIC STRUCTURE FORMATION 47

Figure 3.6: The cosmic large-scale structure as seen with the Millennium Run simulation. The top panels show the dark matter distribution at redshift z =1.4 for a comoving image scale of approximately 500 Mpc (left) and 125Mpc (right) on a side. The same regions at redshift z = 0 are illustrated in the bottom panels. Note that the cosmic web is basically in place at z> 1, but the nodes, i.e. the clusters, still actively accrete large amounts of matter until the present. Credits: V. Springel16.

to allow observations of a large amount of structure17. A sensible choice for fixing the filtering scale is R 8 h−1Mpc, which was motivated by results of early galaxy surveys −≡1 finding δgal(R =8 h Mpc) δNgal/Ngal δM/M 1, i.e. the variance of galaxy counts in spatial bins of radius 8 h≃−1Mpc is similar≃ to the≃ mean value. The mass scale of this 14 filtering length is M 5 10 M⊙ and corresponds to the collapsed region of a massive galaxy cluster. The mass≃ × variance of order unity on the 8 h−1Mpc scale also marks the transition region from the linear regime with δ < 1, i.e. L > 8 h−1Mpc, to the non-linear regime at L 8 h−1Mpc. ∼ ∼ ≪ 18 In short, σ8 is the average fluctuation amplitude of the linearly evolved DM,lin (Equs. 3.31 & 3.32), present day dark matter density contrast field δ8 (x, t0) smoothed with a top-hat filter of comoving radius R = 8 h−1Mpc. The formal definition and its

17The homogeneity of the Universe on very large scales decreases the mass variance with increasing 2/3 filtering length R. For large R and the use of a spherical top-hat filter one obtains σM δM/M M − . 18Meaning the root-mean-square (rms) or standard deviation of the Gaussian field.≃ ∝ 48 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

connection to the power spectrum is then given by

2 2 2 dkk DM 2 − ˜ − σ8 = δ8 h 1Mpc 0 =4π 3 P0 (k) W8 h 1Mpc(k) . (3.40) h i Z (2π) The single dimensionless number σ8 hence determines the height of density peaks and consequently the object abundance. The discussed galaxy cluster mass function (Equs. 3.38 & 3.39) provides a direct measure of σ8. (Nevertheless, σ8 is probably the least securely determined prime cosmological parameter.) The second important parameter related to the large-scale structure in the Universe is nS the power law index nS of the primordial (i.e. initial) power spectrum Pi(k) k . This index is observationally confirmed (Spergel et al., 2007) to be very close to∝ the ‘scale- HZ 1 free’ Harrison-Zeldovich spectrum Pi (k) k (i.e. nS = 1), which is also predicted by inflationary theories. ∝ From the known evolution of the density fluctuation field δ(t) D (t) (Equ. 3.31) and ∝ + the definition of the power spectrum (Equ. 3.35), the time evolution of P (k, t) starting with the initial fluctuation spectrum seeded after inflation is then given by

P (k, t)= D2 (t) Ak1 T 2(k, t) . (3.41) + · · The a priori unknown amplitude A of the power spectrum (today) is directly linked to σ8 as an alternative normalization method (cluster normalization). T (k) is the so-called transfer function which is an extra term containing all evolutionary effects that alter the original linear form of the primordial seed power spectrum. In particular, the transfer function incorporates important scale-imprinting effects on the initially scale-invariant power spectrum. The first imprinted physical scale with prime influence on the present day overall shape H of P (k) is the (particle) horizon at the epoch of matter-radiation equality Dequ at redshift zequ. This characteristic length scale represents the distance over which causal influence can propagate during the radiation dominated era of the Universe and is given by c 1 DH = 16(Ω h2)−1 Mpc 100 Mpc . (3.42) equ √ m 2 H0 (1 + zeq) Ωm ≃ ∼ During the radiation dominatedq early epoch with a coupled photon-baryon fluid, the pressure term in Equ. 3.30 becomes important and causes an oscillatory behavior of the fluctua- H < H tions inside the horizon Dequ. Fluctuation growth on scales L Dequ is therefore effectively ∼ H suppressed, and the structure growth can only proceed on super-horizon scales L > Dequ beyond the causal influence of the radiation pressure. For the following arguments, we only consider the (cold) dark matter component as the dominant contribution to the matter density. The decisive criterion for the growth of perturbation modes with wavenumber k =2π/L during the rapid expansion phase of the radiation-dominated early epoch is determined by whether or not the length scale of the mode is contained within the expanding horizon DH(z) at redshift z. Modes that H are outside this region of causal influence kgrow(z) < 2π/D (z) do not ‘feel’ the radiation 3.2. COSMIC STRUCTURE FORMATION 49

pressure implying that the modes grow according to the linear expectations of Equ. 3.41 with T (k) = 1. Once the expanding causality horizon encompasses perturbation modes H ksuppr(z) > 2π/D (z), the growth suppression sets in and the perturbation amplitudes essentially remain constant. For the growth evolution in the radiation dominated phase starting after inflation and ending with the epoch of matter radiation equality at zequ 3500, we obtain the following scenario: (i) Initially the primordial Harrison-Zeldovich spectrum∼ evolves homogeneously 2 as the modes on all length scales grow proportional to D+(t), i.e. only the amplitude of the linear power spectrum increases with time. (ii) Once the length scale of a perturbation H mode enters the particle horizon at zenter, i.e. kenter(zenter)=2π/D (zenter), further growth is suppressed and the amplitude of the mode halts at the value of horizon entering. (iii) This process proceeds from the smallest scales, i.e. large wavenumbers k, to larger scales as the horizon expands with time, which implies that the structure growth suppression first influences small regions and over time gradually halts the growth for modes of increasing scales. Upon horizon entering, the spectral power density of the corresponding perturbation mode P (kenter) will be held constant, while the modes of smaller wavenumbers continue to grow. The overall shape of the power spectrum P (k, t) will hence bend over towards lower amplitudes for modes with k >kenter(z). (iv) After matter-radiation equality zeq, the matter density dominates the dynamics∼ of the Universe. The photon-baryon fluid decou- ples, resulting in a vanishing pressure for the baryons and an associated resumed structure growth on all scales. The largest length scale which experiences a growth suppression is H thus the introduced horizon scale at equality Deq, which defines the break point keq of the present power spectrum. This horizon-crossing growth suppression mechanism in the early Universe completely fixes the overall shape of the present day cold dark matter power spectrum PCDM(k, t0) H imprinting the characteristic length scale Deq as a break point. For small wavenumbers k 2π/DH the form of the primordial power spectrum is conserved with P (k ≪ eq CDM ≪ keq, t0)=Ak, i.e. T (k) 1. Large wavenumbers with k keq have experienced an extended ≈ ≫ H −2 early period of growth suppression resulting in a transfer function T (k) (Deq k) and −3 ≈ consequently a power spectrum of the form PCDM(k keq, t0) k . The break point k =2π/DH (Ω h2) Mpc−1, i.e. the maximum of the≫ power spectrum,∝ is sensitive to the eq eq ∝ m matter density Ωm and moves towards higher wavenumbers with increasing matter content, reflecting the earlier epoch of matter-radiation equality. In principle, this result compresses the statistical information present in the simulated large-scale dark matter distribution shown in Fig. 3.6. The locally observed galaxy cluster power spectrum is shown in the next Chapter in Fig. 4.2 (upper curve in right panel). The broad maximum of the power spectrum is visible but will only be well constrained with the extended sample of local X-ray cluster surveys currently in preparation. Note that the well-fit P (k) k−3 behavior at large wavenumbers confirms the cold dark matter ∝ paradigm. Hot (HDM) or warm (WDM) forms of dark matter would exhibit a different power spectrum shape, as small scale perturbations are additionally damped by diffusion processes of the fast particles (‘free streaming’). The final ingredient to fully understand the observed cluster power spectrum of Fig. 4.2 50 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

and its vertical offset to the galaxy power spectrum is the concept of biasing (Kaiser, 1984). So far we have mostly related the discussion to the cold dark matter power spectrum PCDM(k), which is not directly accessible to observations. Different classes of observable objects could have systematic offsets of their density field with respect to the dark matter density contrast δ =b δ . The pre-factor b is the bias of the object class which can object · CDM be determined from first principles. The origin of galaxy clusters from the highest peaks in the initial density contrast field results in a large bias factor of b 2.6, i.e. clusters have an amplified19 power spectrum P (k)= b2 P (k) 6.8 P ≃(k) 6.8 P (k). This cluster · CDM ≃ · CDM ∼ · galaxy greatly enhanced clustering strength can be understood in terms of the underlying long- wavelength fluctuation modes of the dark matter density field. The probability for a cluster- scale density perturbation to reach the critical collapse threshold value δc is highly increased if this region is found on top of a larger perturbation mode with small wavenumber. On the other hand, the probability is suppressed if the cluster-scale perturbation coincides with a minimum of the underlying long-wavelength mode. The biasing of different object classes is hence expected to increase with object mass. Clusters in this respect are amplified probes and can thus yield valuable results with a lower number of objects compared to galaxies. The second important imprinted physical scale in the transfer function T (k) is the sound horizon at matter-radiation decoupling. We are now considering the baryon component of the matter density which implies that the expected second order baryonic effects are about an order of magnitude smaller than those due to the dark matter. Nevertheless, the theory is well understood (e.g. Hu & Sugiyama, 1996; Eisenstein & Hu, 1998), and observations of the so-called baryonic acoustic oscillations (BAOs) bear great promise to yield valuable results for the Dark Energy studies over the next two decades. The inclusion of baryons during the strongly coupled baryon-photon fluid phase before recombination at redshift zrec requires the consideration of the last oscillation term in Equ. 3.30. For small scales inside the sound horizon (SH) DSH(z) the relativistic photon- baryon system behaves like a damped, forced oscillator where the self-gravitation of matter overdensities acts as driving force and the photon induced pressure as counter action. In this relativistic fluid sound waves, i.e. pressure and density perturbations, can propagate at the sound speed of c = ∂p/∂ρ c/√3. At the epoch of hydrogen recombination s ≃ at z 1 100, photons andq baryons decouple, which implies that pressure waves cease rec ≈ to expand and are frozen into the (baryonic) matter distribution. The longest baryonic acoustic oscillation wavelength, i.e. the first harmonic, is hence the comoving distance a sound wave could have propagated before recombination and is given by the sound horizon at zrec (Hu, 2004)

DSH 147 (Ω h2/0.13)−0.25(Ω h2/0.023)−0.08 Mpc = 147.8 2.6 Mpc . (3.43) rec ≃ · m b ± The numeric value was measured with WMAP (Spergel et al., 2007) with about 2% accuracy and is expected to be determined on the percent level within the next few years.

19The biasing factor depends on the typical cluster mass in a survey. The given numeric values are taken from the REFLEX survey (see Sect. 4.2). 3.3. COSMOLOGICAL TESTS WITH GALAXY CLUSTERS 51

The first harmonic acoustic wave hence imprints an overdensity in the matter distribution which will be detectable as a peak, implying enhanced structure, in the 2-point correlation SH function ξ(r) at scale Drec or as a modulation peak in the power spectrum at the corresponding wavenumber. Higher order harmonics give rise to additional acoustic peaks at shorter wavelengths resulting in a characteristic baryonic acoustic oscillation pattern in the power spectrum. SH The imprinted sound horizon of Drec 148 Mpc acts as a high-precision comoving ‘standard ruler’. In principle, any object class≃ can be used to probe the baryonic acoustic oscillations and to determine their locations in the power spectrum as a function of redshift. However, as second order effect, BAO measurements are extremely difficult and require high-precision measurements of P (k). It is also expected that the oscillations are damped or washed out at low redshift due to the onset of non-linear effects. The detection of the first acoustic peak in the correlation function of luminous red SDSS galaxies at z 0.35 has been reported (Eisenstein et al., 2005). The measurement of baryonic acoustic∼ oscillations in power spectra of galaxy clusters provides a potentially powerful cosmological tool for the future (e.g. Angulo et al., 2005) but requires cluster samples of the order of ten thousand.

3.3 Cosmological Tests with Galaxy Clusters

In the last two sections some important aspects of the close interplay between the observable properties of galaxy clusters and the underlying global geometry and structure formation process in the Universe were introduced. This section will shortly summarize a number of cosmological tests that use galaxy clusters as probes and have been successfully applied in the past or will become feasible within the next decade. For the dawning era of precision cosmology it will be crucial to measure cosmological parameters with tests sensitive to the global geometry of the homogeneous Universe and methods probing the structure evolution of the inhomogeneous Universe. The very different physical nature of the methods and their underlying theoretical frameworks will allow important consistency cross-checks of the emerging cosmological concordance model. Galaxy clusters in this respect enable several independent cosmological tests for both fundamental aspects which qualifies the cluster population as one of the most versatile cosmological probes. The majority of cosmological tests can be reduced to the measurement of effectively four cosmic evolution functions for which the dependence on the underlying cosmological parameters have been discussed: (i) the cosmic expansion history H(z) (Equ. 3.16), (ii) cosmic distances D(z) (Equs. 3.19, 3.20, 3.21), (iii) cosmic age t(z) (Equ. 3.28), and (iv) the linear growth factor D+(z) (Equ. 3.32). The first three methods are linked to the global geometry and probe the homogeneous Universe, the last one traces the structure growth of the inhomogeneous Universe. (1) Cluster Mass Function: The cluster abundance of the local Universe mainly depends on the matter density Ωm and the amplitude of the density fluctuation field 52 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

σ8. The mass function is related to the statistics of rare high amplitude peaks in the primordial matter distribution. For a direct comparison to the observed cluster population, the predicted mass function n(M) is converted to the measured X-ray luminosity function (XLF) via φ(LX)= n[M(LX)] dM/dLX . The X-ray luminosity function at z 0 has been well established (Böhringer· et al., 2002a) and serves as ∼ local reference for evolution measurements. The local XLF is degenerate concerning ∼0.5 the parameter combination σ8 Ωm , i.e. the effect of a larger fluctuation amplitude can be compensated by a lower· matter density. The degeneracy can be broken in combination with the power spectrum or by measuring the redshift evolution in the next test.

(2) Number Density Evolution: The redshift evolution of the cluster abundance out to z 1.5 will be the main cosmological test of the final XDCP sample. The evolu- ∼ tion of the cluster mass function n(M, z) combines two tests simultaneously: (i) the comoving volume elements dV D2/H(z), governed by the expansion history, and com ∝ (ii) the growth of the mass fluctuation spectrum driven by D+(z). The cosmological test also becomes increasingly sensitive to variations of the Dark Energy parameters ΩDE and w at redshifts above z > 0.8. Up to now, the changing demographics of the entire cluster population has∼ been investigated out to intermediate redshifts of z 0.8 by Mullis et al. (2004) and Borgani et al.(2000) (see Fig. 4.3), but only weak evolutionary∼ effects have been found. The redshift at which the cluster number density decreases significantly is yet to be determined and will be one of the goals for XDCP.

(3) Amplitude and Shape of Power Spectrum: The clustering properties and the general shape of the cluster power spectrum depend on the matter density Ωm and the amplitude of the dark matter ﬂuctuation spectrum σ8 as discussed in Sect. 3.2.2. Reliable measurements require large area contiguous surveys such as the ROSAT All- Sky Survey. The currently best measured cluster power spectrum has been obtained with the REFLEX survey (see Fig. 4.2) which will soon be improved upon with the extended NORAS 2 and REFLEX 2 samples (see Sect. 4.2). Future X-ray surveys such as eROSITA (see Sect. 12.2) will be able to measure the power spectrum as a function of redshift and will thus directly probe the evolution of the growth function D+(z).

(4) Baryonic Acoustic Oscillations: Measurements of the locations of the acoustic peaks in the cluster power spectrum as a function of redshift will be one of the main aims and challenges of the eROSITA mission (Sect. 12.2) and will require a total of 50 000–100 000 clusters with approximate redshifts. This cosmological test, however, is one of the cleanest methods to probe Dark Energy since it is least aﬀected by systematic uncertainties. It is based on the well-understood ‘standard ruler’ im- SH printed by the sound horizon at the epoch of last scattering Drec . Measurements of the transverse ‘baryonic wiggles’ probe the angular size distance dang(z) (Equ. 3.21) 3.3. COSMOLOGICAL TESTS WITH GALAXY CLUSTERS 53

while measurements in the radial direction trace the expansion history H(z) through Equ. 3.19.

(5) Cluster Baryon Mass Fraction: This conceptually simple and attractive test rests upon the assumption that clusters, as objects that have collapsed from a region of 10 h−1 Mpc, represent a fair accounting of the relative matter content in the Universe, i.e. that their baryon mass fraction fb is representative. The cluster gas mass fraction fgas can be well measured with X-ray observations. With corrections for the galaxy mass in clusters, one obtains the universal total cluster baryon fraction which is asymptotically approaching values of f 0.12–0.14 h 3/2 towards the cluster b ∼ 70 outskirts. If the baryon content Ωb of the Universe is taken as prior from cosmic nucleosynthesis calculations, the total matter density can be determined by Ωm = Ωb/fb (e.g. Allen et al., 2003).

(6) Evolving Gas Mass Fraction: This method is a generalization of test (5) and uses the cluster gas mass fraction fgas as ‘standard ruler’ to probe the global geometry of the Universe (Sasaki, 1996). The additional assumption is that the gas mass fraction is universal at all redshifts f (z) const. Since measurements of the gas mass gas ≈ (M h−5/2 d5/2 ) and total mass (M h−1 d ) obey different scaling relations gas ∝ ang tot ∝ ang with distance, the resulting observed gas mass fraction scales with f (z) d3/2 (z). gas ∝ ang The measured gas mass fraction hence depends on the assumed angular diameter distance to the cluster, and will deviate from the (assumed) universal value f¯gas(z = 0), if the reference cosmology is incorrect. This geometric test can constrain the parameters Ωm, ΩDE, and w and has been applied by Allen et al. (2004) based on detailed X-ray observations of 26 dynamically relaxed clusters at z < 0.9. ∼ (7) Absolute Distance Measurements: When X-ray and SZE observations are combined, the absolute distance to a cluster can be determined, allowing the measurement of the Hubble constant H0. The SZ effect (see Sect. 2.3.6) yields the observed temperature decrement ∆T n T dl n T R , and the observation of the thermal ∝ e e ∝ 0 0 cl X-ray emission (see Sect. 2.3.1) results in the flux S n2 Λ(T ) dl n2 R , where R X ∝ e e ∝ 0 cl Rcl is the clusters radius, n0 the gas density, and T0 the gasR temperature. Elimination of n0 results in an expression for the physical cluster radius Rcl, which can be related to the observed angular diameter θcl to yield the angular diameter distance via

2 Rcl ∆T 1 c z dang(z)= = 2 const . (3.44) θcl SX T0 θcl · ≈ H0

The last step is a low redshift approximation for H0, which can be generalized for more distant clusters to H(z). Achievable accuracies for H0 based on individual cluster measurements are about 20 km s−1Mpc−1. ± (8) Cluster Number Counts: Cumulative cluster number counts provide a model independent test since only X-ray ﬂux measurements are needed. Generalizing Equ. 3.26 54 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES

for the cluster population yields the number of clusters per steradian brighter than ﬂux S (e.g. Borgani, 2006) as

∞ cD2(z) ∞ dM N(>S)= dz dS n[M(S, z); z] . (3.45) Z0 H(z) · ZS dS Here the first integral accounts for the comoving volume per steradian (Equ. 3.25) and the second term yields the observable clusters in this volume at flux S. Deviation of the Euclidean slope of Equ. 3.26 are hence related to (i) the flux limit, (ii) evolution effects of the cluster population, and (iii) cosmological distances. A compilation of different log N–log S functions is shown in Fig. 4.6. The observed slope on the bright end starts with the expected Euclidian value 1.5 and then flattens to a value of 1 ≃ towards faint fluxes (Rosati et al., 2002). Cluster number counts are nowadays mainly used as a consistency check with other surveys and for an estimation of the survey flux limit.

(9) BCG Hubble Diagram: The brightest cluster galaxies exhibit a remarkably low scatter of about 0.3 mag in their average absolute magnitude M¯ , which qualiﬁes ± BCG them as decent ‘standard candles’, similar to supernovae type Ia. With the observed apparent BCG magnitudes mBCG of a given ﬁlter, we arrive at the mBCG–log(z) Hubble diagram by combining Equ. 3.23 & 3.24

m (z)= M + K (z)+25+5log(d [Mpc]) 5log h BCG BCG BCG lum − 70 −1 z ′ c h70 dz = MBCG +KBCG(z)+25+5log [Mpc] +5log (1+z) ′ H0 ! Z0 E(z )! z dz′ = MBCG + KBCG(z) 5log h70 + 43.16 + 5log(1+z)+5log ′ (3.46) − Z0 E(z )!

Here KBCG(z) is the redshift dependent K-correction and in the second line the 20 luminosity distance dlum was substituted with Equ. 3.20. Assuming a concordance model cosmology and considering redshift dependent terms up to an order of (z2) O in Equ 3.46, we obtain the low-redshift approximation

m (z) M + K (z)+43.16+5log z +1.68 z , (3.47) BCG ≈ BCG BCG · with the expected linear m(z) log z behavior for the local approximation z 1. ∝ ≪ Variations of the Hubble constant H0 result in a constant vertical oﬀset, whereas the sensitivity on the cosmological parameters Ωm, ΩDE, and w change the form of the Hubble relation at higher redshifts (last term in Equ 3.46). Based on this eﬀect, SNe Ia measurements have established the existence of a non-vanishing Dark

20Assuming a ﬂat geometry, i.e. k =0. 3.3. COSMOLOGICAL TESTS WITH GALAXY CLUSTERS 55

Cosmological Test Eﬀective Test No Parameters

1 cluster mass function n(M) D+ tens Ωm, σ8 2 number density evolution n(M, z) D+(z), Vcom(z) tens Ωm,σ8,ΩΛ,w 3 cluster power spectrum P (k) D+ hundreds Ωm, σ8 4 baryonic acoustic oscillations dang(z), H(z) ten thousands Ωm, ΩΛ, w 5 cluster baryon fraction fb Ωb/Ωm few Ωm 6 evolving gas mass fraction fgas(z) dang(z) tens Ωm, ΩΛ, w 7 absolute distances dang(z) few H0 8 cluster number counts N(>S) D+ hundreds Ωm, σ8 9 BCG Hubble diagram mBCG(z) dlum(z) tens Ωm, ΩΛ

Table 3.2: Cosmological tests using X-ray galaxy clusters as probes. The table summarizes the discussed methods, the eﬀectively measured cosmic property, lower limits on the number of objects required, and the cosmological parameters for which the method is most sensitive to. The ﬁrst six tests have the potential to provide competitive parameter constraints, while the last three are to be seen as consistency checks without strong constraining power.

Energy component (Perlmutter et al., 1999). The BCG Hubble diagram has been applied for cosmological studies by Collins & Mann (1998) and by Aragón-Salamanca, Baugh & Kauffmann (1998) using clusters at z< 1. With the concordance model in place, the BCG Hubble diagram is nowadays more suitable to probe evolutionary effects in the brightest cluster galaxy population as will be shown in Sect. 11.1.

The cosmological tests using X-ray luminous galaxy clusters as probes are summarized in Tab. 3.2. The methods can be loosely classiﬁed into two categories: (A) tests 1–6 having the potential to yield competitive constraints on cosmological parameters, and (B) tests 7–9 which are more to be considered as consistency checks, either for the underlying model (7), survey cross-comparisons (8), or to disentangle evolutionary eﬀects (9). For distant cluster cosmology, the uncertainties in the mass calibration derived from observed properties and the redshift evolution of the scaling relations pose the largest current limitations. Improving these uncertainties and characterizing z > 1 clusters as cosmological probes are important goals for the XDCP survey. The main cosmological application of the XDCP sample will be test 2, which is expected to yield competitive parameter constraints 4–5 years from now. For tests 6 & 7, XDCP can provide additional high-z clusters in order to increase the redshift leverage. Methods 8 & 9 are applicable even for an uncompleted survey and are presented in Sect. 9.1.2 & 11.1. 56 CHAPTER 3. GALAXY CLUSTERS AS COSMOLOGICAL PROBES Chapter 4

X-ray Cluster Surveys

This last introductory chapter will shortly summarize some of the major achievements of galaxy cluster X-ray surveys. In addition, the state-of-the-art of distant X-ray luminous clusters at the start of this thesis will be discussed.

4.1 X-ray Selection

Cosmological evolution studies of the galaxy cluster population impose three essential requirements on any suitable survey sample (e.g. Rosati, Borgani & Norman, 2002): 1. An efficient method to find clusters over a wide redshift range; 2. An observable estimator of cluster mass; 3. A method to evaluate the survey volume within which clusters are found. Finding distant clusters is the main aim of this thesis and will be discussed in detail in the next four chapters. A suitable observable proxy for the cluster mass has to be calibrated with local surveys and is shown in the next section. The survey volume Vmax or equivalently 1 the selection function Ssky[f(L, z)] , i.e. the effective sky coverage as a function of limiting flux, are essential to determine the comoving density of clusters and hence to relate the observations to model predictions. For a strictly flux-limited X-ray survey with limit flim and corresponding sky coverage S[flim], the comoving maximum search volume Vmax(L) for objects with luminosity L is straightforward to calculate using Equ. 3.25. The maximum redshift zmax is obtained from the luminosity distance (Equ. 3.20) by solving d (z ) L/(4πf ) for z . The lum max ≡ lim max search volume (see Fig. 3.2) is then given by q

z 2 max dlum(z) c dz Vmax(L)= Vcom[zmax(flim, L)] = Ssky[flim] . (4.1) Z0 1+ z ! H(z) 1 Here the ﬂux is denoted with f and the sky coverage in units of steradian with Ssky, not to be confused with the ﬂux S often used as a notation for number counts.

57 58 CHAPTER 4. X-RAY CLUSTER SURVEYS

For an inhomogeneous sensitivity distribution, as typical for serendipitous surveys, the sky coverage2 rises gradually over an extended flux range from the most sensitive regions with limit flow to the flux fhigh covering the full survey area. In this case, the search volume is obtained by a second integration over the flux limit and proper accounting of the additional survey area dS/df at higher fluxes

f f z (f) 2 high dVcom[zmax(f, L)] high max dSsky[f] dlum(z) c dz Vmax(L)= df = df . (4.2) Zflow df Zflow Z0 df 1+ z ! H(z) In principle, galaxy clusters can be systematically selected in several independent ways: (i) Optical/NIR selection based on the cluster signature of a spatial galaxy overdensity (and clustering in color or redshift space), (ii) X-ray detection of the extended ICM emission, and for future surveys (iii) selection based on the SZE, or (iv) the weak gravitational lensing signature of clusters. The main important advantages of X-ray galaxy cluster surveys can be summarized as follows: Bright thermal X-ray emission is only observed for well evolved clusters with a • deep gravitational well (Sect. 2.3); The X-ray emission is highly peaked which minimizes projection eﬀects • (Sect. 2.3); The X-ray luminosity is tightly correlated with the gravitational mass and is provid- • ing a reliable mass proxy based on the survey data (Sect. 4.2); The X-ray selection can be accurately assessed allowing a simple evaluation of the • survey volume (this section); Cosmological applications with X-ray clusters can build upon several decades of • survey experience (Chaps. 1 & 4).

4.2 Local Cluster Surveys

The success story of X-ray surveys for clusters of galaxies is largely built upon the German- American ROSAT (ROentgenSATellit) mission from 1990–1999. The X-ray imaging capabilities of its main instrument, the Position Sensitive Proportional Counter (PSPC), allowed for the first time a cluster selection based on source extent, in particular for the mission phase of pointed observations after the completion of the ROSAT All-Sky Survey (RASS). The central 0.2 deg2 of the PSPC field-of-view achieved a point spread function with a FWHM resolution of 30′′–45′′ enabling the resolved detection of extended emission of clusters with fiducial core radii of 250 h−1 kpc out to z 1 (e.g. Rosati, Borgani & Nor- man, 2002). Considering that the extragalactic AGN point∼ source population outnumbers clusters by approximately a factor of 10:1 at faint flux levels (f 10−14 erg s−1cm−2), the X ∼ selection based on the extent criterion greatly enhances the cluster survey efficiencies. 2Additional correction terms might be necessary, e.g. to take the detection efficiency as a function of cluster core radius or off-axis angle into account (see Sect. 9.1.3). 4.2. LOCAL CLUSTER SURVEYS 59

XDCP

Figure 4.1: Comparison of X-ray cluster surveys over the past two decades. The approximate survey ﬂux limits and solid angles are plotted for diﬀerent contiguous area surveys (light shaded circles) and serendipitous surveys (dark circles). The position of the XDCP survey has been added. Plot adapted from Rosati, Borgani & Norman (2002).

Two basic types of X-ray surveys can be distinguished. (i) Contiguous sky area surveys covering a single connected region, and (ii) serendipitous surveys making use of the ensemble of sky patches of numerous individual pointed observations. Both survey types have their merits and strengths for different science applications. Contiguous surveys, in particular with all-sky coverage, span large solid angles with typically relatively bright flux limits. They are well-suited to detect the rarest, most luminous systems in the large survey volume and to investigate their clustering properties. Serendipitous surveys, on the other hand, are typically sensitive to significantly fainter flux limits, thus providing complementary information on more common systems with lower luminosities and more distant objects. Samples compiled from this survey type are particularly fit for cluster evolution studies. Figure 4.1 provides an overview of the major X-ray cluster projects of the last two decades. Contiguous area surveys are indicated by the light shaded symbols, whereas serendipitous surveys are represented by dark circles. The approximate location of the XMM-Newton Distant Cluster Project in the solid angle versus flux limit plane has been added in blue. First, we will have a short look at the major all-sky survey samples providing a local cluster census and reference system. 60 CHAPTER 4. X-RAY CLUSTER SURVEYS

Figure 4.2: Important results of local cluster surveys. Left: The empirical LX–M relation of Reiprich & Böhringer (2002) based on the HIFLUGCS sample of the 63 brightest galaxy clusters in the sky (solid line) and an extended sample of 106 objects (dashed line). Right: Galaxy cluster power spectrum from the REFLEX survey (top filled symbols) in comparison to a galaxy power spectrum (bottom open symbols). The power spectra amplitude of the two object classes differ by the biasing factor b2 6.8. Plot from Schuecker et al. (2001). ≃

Figure 4.3: X-ray luminosity functions (XLF) as determined by diﬀerent cluster surveys. Left: The local (z< 0.3) X-ray luminosity functions of eight ﬂux-limited surveys show excellent agreement and serve as the no-evolution baseline for comparisons with distant cluster samples. Right: XLFs derived from cluster samples of intermediate redshifts (0.25

Figure 4.4: The NORAS2 cluster of galaxies RXC J0834.9+5534 at z = 0.239 as a typical example of a low-redshift object detected in the ROSAT All-Sky Survey. The red X-ray contours are shown on top of the 12′ 12′ op- × tical DSS image. The circles indicate the locations of spectroscopically conﬁrmed cluster galaxies of early type (red), late type (blue), and Seyfert type 1 (yellow).

Figure 4.5: Typical spectra of galaxy members of the cluster RXC J0834.9+5534 at z =0.239 shown above. Important spectral features are indicated by dashed blue lines with corresponding labels, atmospheric features by red dotted lines. Left: Early type galaxy spectra with their characteristic D4000 break redshifted to about 5 000 A˚ and other labelled absorption features. Right: Late-type galaxy spectrum (top) and a Seyfert type1 spectrum (bottom) with indicated emission lines. 62 CHAPTER 4. X-RAY CLUSTER SURVEYS

HIghest X-ray FLUx Galaxy Cluster Sample (HIFLUGCS): This sample includes the 63 brightest X-ray clusters in the (extragalactic) sky and has a ﬂux limit −11 −1 −2 of fX = 2 10 erg s cm in the ROSAT 0.1–2.4 keV band (Reiprich, 2001). The major× achievement with this brightest cluster sample has been the empirical calibration of the (local) LX–M relation as shown in the left panel of Fig. 4.2 (Reiprich and B¨ohringer, 2002). This relation establishes the crucial mass proxy and hence the link to cosmological applications for large X-ray survey samples.

ROSAT-ESO Flux-Limited X-ray cluster survey (REFLEX): The REFLEX sample (Böhringer et al., 2004) is currently the largest statistically complete published compilation of X-ray clusters. It contains 447 objects in the Southern sky (δ<+2.5◦) down to the flux limit of f =3 10−12 erg s−1cm−2 [0.1–2.4 keV]. REFLEX provides X × the local reference of the luminosity function (left panel of Fig. 4.3) (Böhringer et al., 2002a) and has been used for a wide variety of cosmological studies (e.g. Schuecker et al., 2003a, Schuecker et al., 2003b). A major achievement has been the measurement of the galaxy cluster power spectrum on scales of 15–800 h−1 Mpc as shown in the right panel of Fig. 4.2 (Schuecker et al., 2001). An extension of the survey, RE- FLEX 2, is in preparation and will roughly double the number of objects by lowering the flux limit to f =1.8 10−12 erg s−1cm−2. X × NOrthern ROSAT All-Sky survey (NORAS): NORAS (Böhringer et al., 2000) is the counterpart of REFLEX in the Northern hemisphere containing 378 clusters out to z <0.3. The survey extension NORAS 2 is in preparation. The combined extended samples∼ NORAS 2 and REFLEX 2 with a total of approximately 1 800 X-ray clusters will provide the next milestone for a local power spectrum analysis. This thesis work also included the spectroscopic analysis of 35 clusters for the extended NORAS 2 survey. Figures 4.4 & 4.5 show one example of a low-redshift cluster at z = 0.239 and some typical cluster galaxy spectra as local reference for the discussions on the spectral energy distribution (SED) in Chap. 7 and spectroscopic features (Chap. 8) of distant cluster galaxies.

4.3 Deep Surveys

Three of the major deep ROSAT surveys are introduced and the corresponding cluster samples used as the basis for discussing evolution eﬀects of the X-ray luminosity function.

North Ecliptic Pole survey (NEP): The North ecliptic pole is the area with the deepest eﬀective exposure time of the ROSAT All-Sky Survey owing to the survey scan pattern along great circles perpendicular to the ecliptic. 81 square degrees with the most sensitive coverage around the North pole have been used as the basis for a complete spectroscopic identiﬁcation of all X-ray sources. The North Ecliptic Pole −14 −1 −2 survey (Henry et al., 2006) with a median sensitivity of fX 7 10 erg s cm includes 62 galaxy clusters out to a maximum redshift of z ∼=0×.81 and is the only deep ROSAT survey with a contiguous area coverage. 4.3. DEEP SURVEYS 63

160 deg2 large area survey (160 deg2): The 160 deg2 serendipitious survey (Vikhlinin et al., 1998) with a median ﬂux limit of 1.2 10−13erg s−1cm−2 [0.5–2.0 keV] and × deepest coverage of 4 10−14erg s−1cm−2 contains 201 clusters. The 22 objects at z> 0.5 allow an assessment× of evolution eﬀects out to redshifts z 0.8. This survey has recently been extended to a sky coverage of 400 deg2 (Burenin∼et al., 2007).

ROSAT Deep Cluster Survey (RDCS): The RDCS (Rosati et al., 1998) is the deepest ROSAT survey available and has compiled a sample of 103 clusters out to z =1.27, which marks the ‘ROSAT redshift horizon’. Based on deep PSPC pointed observations, the serendipitous survey covers a sky area of about 50 square degree down to a ﬂux limit of f = 3 10−14 erg s−1cm−2 [0.5-2.0 keV] yielding 26 systems at X × 0.5 1 X-ray clusters. ×

Figure 4.6 shows a compilation of the cumulative cluster number counts (log N–log S) derived for different local and deep surveys. From the deepest flux levels probed by ROSAT, a surface density estimate of about 10 galaxy clusters per square degree is expected at f 1 10−14 erg s−1cm−2 [0.5-2.0 keV]. On the opposite end, i.e. for the HIFLUGCS se- X ∼ × lection criterion, only one cluster per 400 square degree sky area exists. We can now assess the evidence for an evolution of the galaxy cluster X-ray luminosity function (XLF) as established by ROSAT-based surveys. The compilation of Mullis et al. (2004) in Fig. 4.3 summarizes the measurements of the local cluster luminosity function (left panel) and the distant XLF out to z 0.8 (right panel). The differential luminosity ∼ function φ(LX, z) is commonly parameterized in the form

2 −α d N ∗ LX LX dLX φ(LX, z) dLX = (LX,z)= φ ∗ exp ∗ ∗ , (4.3) dV dLX LX ! −LX ! LX

where N denotes the number of clusters of luminosity LX in volume V at redshift z. The Schechter function on the right hand side is parameterized by the characteristic luminosity ∗ ∗ 3 −3 LX, the faint-end slope α, and the normalization φ (in units h Mpc ), which is directly related to the space density of clusters brighter than a minimum luminosity Lmin via ∞ n0 = Lmin φ(L) dL. TheR local X-ray luminosity function shows a remarkable overall agreement between the surveys considering the independent selection techniques and data sets and is hence a robustly determined reference measure for evolution studies. The parameters for the local XLF can be summarized as follows (e.g. B¨ohringer et al., 2001; Rosati, Borgani & Norman, 2002): L∗ 4 1044 erg s−1 [0.5–2.0 keV], α 1.8 (15% variation), and φ∗ X ≃ × ≃ ≃ 2.7 10−7 h3 Mpc−3 (50% variation). × 70 A meaningful parametrization of evolution eﬀects of the XLF (Rosati et al., 2000) is given by

A −α L φ(L, z)= φ0 (1 + z) L exp , (4.4) · −L∗(z) ! 64 CHAPTER 4. X-RAY CLUSTER SURVEYS

Figure 4.6: Compilation of cumulative cluster number counts as a function of X-ray flux for different ROSAT surveys. The log N–log S has been measured over four orders of magnitude in flux. The right-hand scale indicates the total cluster surface density per square degree at a given flux. Plot from Rosati, Borgani, & Norman (2002).

with the free parameters A for the evolution of the local space density normalization φ0, ∗ ∗ B and B for an evolving characteristic luminosity L (z)= L0 (1 + z) . The current data out to z 0.8 show no detectable evolution effects for galaxy clusters with luminosities 44∼ −1 LX < 10 erg s , i.e. the measurements are consistent with A = B = 0 (right panel of ∼ 44 −1 Fig. 4.3). Only at the highest luminosities at LX > 10 erg s a mild but significant deficit in the volume density compared to the local value is confirmed at a confidence level of >3σ (Mullis et al., 2004). In summary, a moderate negative evolution, i.e. objects were rarer at high redshift, has been established for the most luminous, massive clusters out to z 0.8, but the bulk of the cluster population is approximately constant for at least the second∼ half of cosmic time. This means that the epoch of a significant number density change of the overall cluster population is yet to be determined.

4.4 Distant Galaxy Clusters in 2004

The ROSAT era has led to the discovery of five X-ray luminous galaxy clusters at z>1 (see Tab. 9.1), four of them detected within the RDCS survey, and one in the deep Mega-second observation of the Lockman Hole (LH) field (Hashimoto et al., 2005). The ongoing detailed multi-band follow-up observations of these rare systems have revealed rather advanced evolutionary states despite the large lookback time, showing a metal-rich, hot ICM and 4.4. DISTANT GALAXY CLUSTERS IN 2004 65 a red and passively evolving galaxy population. Figure 4.7 displays a multi-wavelength overlay (left panel) and a high-resolution HST image (right panel) of RDCSJ1252.9-2927 at z =1.237, the most massive and best studied z>1 galaxy cluster in 2004. The situation of distant cluster studies at the start of this thesis is summarized in the cone diagram of Fig. 4.8, illustrating the distribution of known X-ray clusters in redshift space. Out to redshifts of z 0.7 (red circles), the cluster population is well-sampled by the local and deep ROSAT∼ surveys. Orange circles represent the small sample of all known X-ray clusters at z > 0.7, about ten in the redshift range 0.7 < z < 1, and the mere five test objects at 1< z 1.27. The yellow star indicates the position∼ of the cluster XMMU J2235.3-2557 in this diagram,≤ which will be presented in detail in Sect. 5.4. The highly active phase of galaxy cluster surveys has continued into the XMM-Newton era. Here we summarize some of the major projects, which will have a significant impact on the study of distant X-ray clusters.

Cosmic Evolution Survey (COSMOS): The 2 deg2 field of the COSMOS survey (Scoville et al., 2007) has obtained deep multi-wavelength coverage with basically all major astronomical facilities. The project is designed to probe the correlated evolution of galaxies, star formation, AGN, dark matter, and the large-scale structure. More than 2 Msec of XMM-Newton data have been obtained (Hasinger et al., 2007) allowing the detection of extended sources down to flux levels of f 10−15 erg s−1cm−2. X ∼ An early set of the COSMOS X-ray data has been used to test the XDCP selection procedure and efficiency (Sect. 6.5). The COSMOS survey has the ability to detect and study dozens of high-redshift groups and clusters (Finoguenov et al., 2007). However, the limited solid angle restricts the survey to the inclusion of a single structure of order 2 1014 M , implying that the high mass end of the XLF cannot be × ⊙ sufficiently probed. XMM Large-Scale Structure survey (XMM-LSS): The XMM-LSS project (Pierre et al., 2006) is designed as a designated multi-wavelength galaxy cluster survey over a fairly large contiguous sky area. However, the large demand of XMM resources has limited the currently available X-ray data to roughly 9 deg2 with a typical depth of 10 ksec. The XDCP sample hence provides a wider and deeper coverage. It has an overlap with XMM-LSS of about 0.8 deg2, which can be used for a survey comparison study. XMM Cluster Survey (XCS): The ambitious XCS (Romer et al., 2001) serendipitous survey aims at the detection of all galaxy clusters in all XMM archive fields. In principle, XCS fully overlaps with the XDCP survey. However, the XCS project is centered on fairly bright cluster sources with a sufficient number of photons for an accurate temperature determination and is not focussed on very distant clusters with their typically low flux and count levels.

3The images can be found at http://chandra.harvard.edu/photo/2004/rdcs1252/rdcs1252 xray opt.jpg and http://chandra.harvard.edu/photo/2004/rdcs1252/rdcs1252 hst optical.jpg. 66 CHAPTER 4. X-RAY CLUSTER SURVEYS

Figure 4.7: Multi-wavelength view of the distant cluster of galaxies RDCS J1252.9-2927 at redshift z = 1.237, the most massive and best studied z> 1 cluster discovered with ROSAT. Left: 2′ 2′ × color composite image of the cluster X-ray emission (purple) overlaid on the optical/near-infrared data showing the red galaxy population. Right: 1′ 1′ high resolution Hubble Space Telescope image of × the cluster core. Images from press release3.

Figure 4.8: Schematic cone diagram of known X-ray luminous galaxy clusters in 2004. The redshift increases from bottom to top, left to right illustrates the right ascension distribution. The red symbols at z< 0.7 represent a small selection of known low-redshift clusters, orange circles at 0.7 z < 1.3 in- ≤ dicate all known ROSAT X-ray clusters∼ in the high-redshift regime. The XDCP cluster XMMU J2235.3-2557 at z 1.4 (see ∼ Sect. 5.4) was the ﬁrst XMM selected system that could break the ‘ROSAT redshift horizon’. Plot from C. Mullis. Chapter 5

The XMM-Newton Distant Cluster Project

This chapter will provide an overview of the XMM-Newton Distant Cluster Project. The principal survey strategy is introduced followed by a discussion of early results and gained experiences of a pilot study.

5.1 Motivation

The main scientific motivations for starting an extensive new serendipitous distant cluster survey have been outlined in the introductory chapters 2–4. Among the vast number of astrophysical and cosmological applications of distant X-ray luminous galaxy clusters, the accurate determination of the cluster number density evolution out to z 1.5 is the prime ∼ goal and science driver of the XDCP. The time window for a ‘new generation’ survey apt for this ambitious task has only opened in the year 2003 owing to the rapid advance of technology over the last few years. The European X-ray satellite XMM-Newton has been providing an ever increasing amount of archive data with unprecedented sensitivity to a depth that will be unrivalled for at least another decade (see Sect. 12.2). These advances in X-ray technology go hand-in-hand with the availability of ground-based 8–10 meter class telescopes, most notably the European Very Large Telescope (VLT), which now allow the spectroscopic confirmation and further detailed studies of cluster galaxies at z>1. As discussed in Chap. 1, the cluster samples associated with the scientific goals of the XDCP can be grouped into three different categories with increasing time line and scope: (1) provide z> 1 galaxy clusters as individual astrophysical laboratories for detailed studies, (2) compile pathfinder samples for the characterization of high-redshift clusters as cosmological probes for large upcoming surveys, and (3) establish a statistically complete cosmological sample for various cosmological tests. The fundamental pre-requisite for compiling these cluster samples is an efficient survey strategy for finding and confirming the rare distant X-ray luminous systems; the basic steps are introduced in the following section.

67 68 CHAPTER 5. THE XMM-NEWTON DISTANT CLUSTER PROJECT

5.2 XDCP Survey Strategy

The XMM-Newton Distant Cluster Project (Böhringer et al., 2005) is based on the following six-step strategy aiming at the efficient selection and confirmation of X-ray luminous z > 1 clusters of galaxies. The intermediate results of the different survey steps are illustrated∼ in Fig. 5.1 using the galaxy cluster XMMU J2235.3-2557 as an example case. Step 1: Detection of serendipitous extended X-ray sources. Deep XMM-Newton archive fields (>10 ksec) at high galactic latitude ( b 20◦) are analyzed for serendip- | |≥ itous extended∼ X-ray sources in the field-of-view. Step 2: Source screening and selection of candidates. The detected extended X- ray sources are visually inspected and their sky positions cross-correlated with available optical data and extragalactic database information. X-ray sources without an identified optical counterpart are selected as distant cluster candidate sources. Step 3: Two-band follow-up imaging: For the candidate targets, follow-up imaging data is obtained in (at least) two filter bands to sufficient depth, which allows a preliminary cluster identification based on a red overdense galaxy population coincident with the extended X-ray source. Step 4: Photometric redshift estimate. The full photometric analysis yields a cluster redshift estimate based on the color of the red-sequence which allows a refined selection of photometrically confirmed z >0.9 candidates. ∼ Step 5: Spectroscopic confirmation. Promising high-redshift cluster candidates are spectroscopically observed for the final cluster confirmation and redshift determination. Step 6: Detailed follow-up studies. For the most interesting z >1 systems, additional detailed imaging and spectroscopic follow-up observations in X-ray,∼ optical, NIR, IR, and the radio regime will be obtained to disclose the high-redshift cluster properties and physics. It should be emphasized at this point that the vast majority of extended X-ray sources (steps 1 & 2) in high galactic latitude fields are expected to originate from galaxy clusters (see e.g. the NEP survey in Sect. 4.3). Due to the sequential dependency of the different survey steps and the increasing demands on observatory resources, the completion rates at an intermediate survey stage will drastically decrease from top to bottom. The first two steps are discussed in detail in Chap. 6 and are based on publicly available data only, hence providing a 100% coverage for the current survey phase. Steps 3 & 4 are presented in Chap. 7 achieving a survey coverage of about 40% for the thesis data. The observational bottleneck is the spectroscopy of step 5, which is introduced in Chap. 8, and which is currently available only for few systems. The final step 6 will not be discussed in this thesis beyond the overview provided in Sect. 5.4 since detailed follow-up data has only been obtained for the cluster XMMU J2235.3-2557. 5.2. XDCP SURVEY STRATEGY 69

Figure 5.1: Schematic steps of the XDCP survey strategy. I: Detection of serendipitous extended X-ray sources in suitable archival XMM-Newton ﬁelds. II: X-ray source screening using available optical and NIR data (DSS). III: Two-band follow-up imaging to verify the presence of a cluster as overdensity of red galaxies (IIIa R-band, IIIb Z-band). IV: Photometric redshift estimation using the color of the red-sequence as distance indicator. V: Spectroscopic conﬁrmation of the cluster redshift. VI: Detailed multi-wavelength follow-up of the most interesting objects. Plots provided by C. Mullis. 70 CHAPTER 5. THE XMM-NEWTON DISTANT CLUSTER PROJECT

5.3 Pilot Study Results

The XMM-Newton Distant Cluster Project started in the year 2003 with a pilot study to test the feasibility and efficiency of the survey strategy. Since this feasibility study and most of the related work began prior to this thesis, the results and main lessons of this initial project phase are summarized without going into the details. 120 archival XMM-Newton pilot study fields were analyzed for serendipitous extended X-ray sources (step 1). From these about 60 distant cluster candidate sources were selected (step 2) and observed with VLT FORS 2 R and Z-band snap-shot imaging (step 3). Based on the R–Z red-sequence analysis, a dozen promising high-redshift candidates with photometric redshift estimates z > 0.9 could be identified (step 4) and have been proposed for spectroscopic confirmation∼ (step 5). For the first discovered high-z cluster (see Sect. 5.4) detailed multi-wavelength follow-up observations have been proposed and obtained (step 6). The most important XDCP pilot study conclusion was the confirmation that high- redshift galaxy clusters can indeed be efficiently selected based on their extended X-ray signature in conjunction with two-band imaging data. The strategy described in Sect. 5.2 could thus be established as the basis for a large systematic distant galaxy cluster archival survey. The early results have been very encouraging and have pointed to the great scientific potential achievable with such a project. However, besides the positive overall observational feedback, the XDCP pilot study has also revealed optimization potentials for most of the survey chain, in particular for the first four process steps.

Step 1: The pilot study X-ray selection of distant cluster candidates was in ef- • fect based on sources with at least 200 detected X-ray photons. For a 10 ksec net XMM exposure, this implies an approximate detection limit of flim 2 10−14 erg s−1cm−2(0.5–2.0 keV), i.e. comparable to the deepest ROSAT surveys≃ (see× Fig. 4.1). In order to achieve signiﬁcant sensitivity improvements compared to existing surveys, the detection of extended sources with less than 100 X-ray photons is a desirable aim. Consequently, the ﬁrst optimization goal is an improved sensitivity for detecting extended X-ray sources.

Step 2: The pilot study returned a large fraction of intermediate-redshift systems at • z 0.3–0.6, which contributed almost 40% to the identified cluster sample but are not ≃ of particular interest for the prime XDCP science objectives. An improved screening could significantly decrease the number of such intermediate-redshift objects and thus increase the z> 1 cluster fraction in the candidate sample. A more efficient low-redshift cluster identification and rejection is the second optimization goal.

Step 3: For several conﬁrmed high-redshift candidates and clusters between z 1.1– • 1.45 the VLT FORS 2 ‘snap-shot imaging’ with net exposure times of 480s≃ in Z (Z (AB) 23.0mag) and 960s in R (R (AB) 24.5 mag) did not allow a secure lim ∼ lim ∼ 5.4. XMMUJ2235.3-2557 71

cluster identification, i.e. the data did not have sufficient depth. The third optimization goal is thus a more complete two-band imaging cluster confirmation at the high redshift end.

Step 4: The R–Z color exhibits large intrinsic redshift uncertainties at z> 1 . The • ﬁnal optimization goals are thus improved and more secure redshift estimates based on two-band imaging data.

The ﬁrst two optimization items dealing with the X-ray selection process will be discussed and implemented in Chap. 6. The latter two concerning the two-band follow-up imaging will be treated in detail in Chap. 7. As the highlight result of the XDCP pilot study, the galaxy cluster XMMU J2235.3-2557 is now introduced.

5.4 XMMUJ2235.3-2557

The XDCP discovery of XMMU J2235.3-2557 was announced1 in March 2005 with the results published in the accompanying paper of Mullis et al. (2005). At the time of publication, XMMU J2235.3-2557, at a redshift of z =1.393, was the most distant galaxy cluster known, and the first XMM-Newton selected system at z> 1. It was also the first XDCP object for which spectroscopic follow-up observations were obtained. The selection procedure for XMMU J2235.3-2557 is shown in Fig. 5.1 as an illustration of the survey strategy steps. The cluster’s extended X-ray emission was detected in a 44 ksec XMM-Newton observation of the Seyfert galaxy NGC 7314 at an off-axis angle of 7.7′ (panel I). The corresponding sky position on the DSS plates is devoid of objects, which classified the source as a distant cluster candidate (panel II). The subsequent R (panel IIIa) and Z-band (panel IIIb) follow-up images reveal a galaxy cluster signature of red galaxies with a significant central overdensity. The color-magnitude diagram (panel IV) exhibits a cluster red-sequence with a color of R–Z 2.1 indicating a high-redshift cluster candidate ∼ at z > 1. Deep VLT-FORS 2 spectroscopy with an exposure time of 4 hours confirmed the cluster at a redshift of z =1.393 based on 12 identified spectroscopic cluster member galaxies. The detailed multi-band follow-up program (panel VI) is discussed below. The cluster properties, as they have been derived from the survey data set (i.e. steps 1–5), are summarized in Tab. 5.1 and a more detailed view of XMMU J2235.3- 2557 is presented in Fig. 5.2. With 280 net source counts in the 0.3–4.5 keV energy range, XMMU J2235.3-2557 is a fairly bright X-ray source which had been serendipitously detected, without optical identification, in a ROSAT pointed observation. The source flux of (2.6 0.2) 10−14 erg s−1cm−2 in the 0.5–2.0 keV band with a corresponding total unabsorbed± flux× of (3.6 0.3) 10−14 erg s−1cm−2 yields a rest-frame luminosity estimate of (3.0 0.2) 1044 h−±2 erg s−×1 (0.5–2.0 keV). Assuming a thermal (MEKAL) ± × 70 1Press release articles can be found under http://www.eso.org/public/outreach/press-rel/ pr-2005/pr-04-05.html or http://www.mpg.de/bilderBerichteDokumente/dokumentation/ pressemitteilungen/2005/pressemitteilung20050228/presselogin. 72 CHAPTER 5. THE XMM-NEWTON DISTANT CLUSTER PROJECT

emission model with a metallicity of 0.3 Z⊙ and a galactic hydrogen absorption column 20 −2 of nH =1.47 10 cm , a spectral temperature fit (lower right panel of Fig. 5.2) with +2.5× 14 kTX =6.0−1.8 keV is obtained. With a mass estimate of 3 10 M⊙ based on the LX–M relation (see Fig. 4.2), XMMU J2235.3-2557 is likely the mos∼ t× massive galaxy cluster known at redshifts z>1. The derived X-ray properties for the cluster are consistent, within errors, with the determined LX–T relations of other surveys as shown in the right central panel of Fig.5.2. The follow-up imaging observations (top panels and central left panel of Fig. 5.2) of XMMU J2235.3-2557 revealed an old and evolved galaxy population in the cluster center already in place at a lookback time of 9 Gyr. The brightest cluster galaxy exhibits an extended surface brightness profile typical for massive cluster cDs and has a spatial offset to the X-ray centroid of 3′′.7, corresponding to 31 kpc at z = 1.393, which might hint at a recent subcluster merger in the Northeast-Southwest direction. The spectroscopic redshift distribution in the bottom left panel of Fig. 5.2 shows the clear redshift peak at z 1.4 with 12 confirmed cluster members (red) and two additional galaxies at slightly ∼ lower redhsift (orange). The preliminary velocity dispersion estimate based on this data is 762 262 km s−1. ± Meanwhile, XMMU J2235.3-2557 has been the subject of an extensive multi-wavelength follow-up program (step 6). The above discussion has been purposely limited to the discovery data set2 since this will be typical in data quality and quantity for routine detections within the XDCP core survey program (step 1–5). As an outlook of the follow-up science for the most interesting XDCP clusters, we will summarize the obtained follow-up observations and the principal science objectives for XMMUJ2235.3-2557. In X-rays, the cluster has been re-observed with XMM-Newton for a nominal total exposure time of 80 ksec complemented by a 200 ksec observation with Chandra. The main science objectives for the high spatial resolution Chandra data are (i) the reliable measurement of the structural parameters free from the potential impact of point sources for an accurate mass determination, (ii) probing small-scale features (e.g. substructure) to provide further insight into the global state of the system, and (iii) to test for the presence of a cool core. The enhanced spectroscopic performance of XMM-Newton in addition allows (iv) an ICM temperature determination to better than 10%, (v) precise measurements of the X-ray surface brightness, gas mass, and gravitational mass profiles, (vi) the first ICM metallicity evaluation and detection of the iron line at this redshift, and (vii) test the predicted evolving scaling relation (e.g. LX–T , M–T ) and self-similarity of the surface brightness and entropy structure. With all this at hand, XMMU J2235.3-2557 can be (viii) established as a ‘standard candle’ at z 1.4 and used for the cosmological ∼ baryon fraction test of Sect. 3.3. With the additional VLT FORS 2 spectroscopy (ix) more cluster member galaxies can be identified for an improved velocity dispersion measurement and reference data set for photometric studies. The imaging data baseline has been broadened to the R, I, Z, J, H, Ks,

2All preliminary cluster properties have been conﬁrmed (within the stated errors) by the deeper follow- up observations. 5.4. XMMUJ2235.3-2557 73

X-ray Properties Optical Properties −1 −2 −14 fX [erg s cm ] (2.6 0.2) 10 spectroscopic members 12 −2 −1 ± × 44 LX [h70 erg s ] (3.0 0.2) 10 z¯ 1.393 ±+2.5 × −1 kTX 6.0−1.8 keV velocity dispersion 762 262 km s M 3 1014 M color of red-sequence R–Z± 2.1 X ∼ × ⊙ ∼ X-ray photons 280 BCG type cD-like ∼ core radius 140 kpc BCG X-ray centroid oﬀset 31 kpc ∼ Table 5.1: X-ray and optical properties of galaxy cluster XMMU J2235.3-2557 as derived from the initial data set. The unabsorbed ﬂux fX and the rest-frame X-ray luminosity LX are given for the 0.5–2.0 keV band.

3.6 µm, and 4.5 µm-bands, i.e. from 0.6–4.5 µm corresponding to a rest-frame photometric coverage from the UV to the NIR. Deep I, Z, and H-band coverage has been obtained with the Hubble Space Telescope ACS and NICMOS cameras (>30hours in total), J (2700s) and Ks-band (3 600 s) data with VLT ISAAC, and the mid-infrared 3.6 µm and 4.5 µm- bands with the Spitzer observatory (4 hours in total). This eight-band data set allows (x) detailed SED fitting of the spectroscopic members to yield formation ages, crude star formation histories and stellar masses, (xi) an accurate morphology mix determination based on the high-resolution HST data to probe whether the galaxy Hubble sequence is already in place, and (xii) a comparison of the ages and stellar masses of early-type cluster galaxies with the field elliptical population at similar redshifts to probe the expected accelerated galaxy evolution in high-density environments. (xiii) The accurate measurement of the normalization, scatter, and slope of the Z–Ks red-sequence will accurately constrain the formation epoch and can probe evolutionary trends compared to lower redshift studies. (xiv) The construction of the rest-frame NIR luminosity function out to M*+3 can shed light on the role of merging and stellar mass assembly of galaxies, since the faint-end slope is expected to steepen at high redshift as the cluster galaxies in hierarchical models break up into their progenitors. Additional science objectives mentioned here are (xv) fundamental plane (FP) studies of the cluster ellipticals and (xvi) an investigation of the cluster mass-to-light ratio in combination with the available X-ray data. In summary, the XDCP discovery of the massive galaxy cluster XMMU J2235.3-2557 at z =1.393 has triggered an extensive follow-up program to address a number of open questions within the cluster and galaxy evolution scenarios to be discussed in detail in forthcoming publications. At this point it should be emphasized again that the search for such rare massive high-redshift systems is best performed within the scope of large- area serendipitious surveys covering several dozen square degrees. For the discussions in the upcoming chapters one should keep in mind that less than one out of a hundred serendipitous extended X-ray sources will be of similar cluster quality and scientific impact. XMMU J2235.3-2557 has thus marked a very positive XDCP start and has encouraged the extended systematic survey for this thesis work. 74 CHAPTER 5. THE XMM-NEWTON DISTANT CLUSTER PROJECT

Figure 5.2: A detailed view of galaxy cluster XMMU J2235.3-2557 at z = 1.393. Top left: VLT- FORS 2 R-band image (1 140 s) with X-ray contours overlaid in blue. Top right: 1.5′ 1.5′ zoom on × a 3 600 s VLT-Isaac Ks-band image. The colored objects correspond to the red and orange peaks of the redshift histogram in the bottom left panel. Center left: R+Z+Ks pseudo-color composite with field size 2.5′ 2.5′. Center right: XMMU J2235.3-2557 (red filled diamond) in comparison to L -T × X relations as derived for different surveys (Ettori et al., 2004). Bottom right: Spectral X-ray energy distribution of XMMU J2235.3-2557 (vertical bars) in the 0.5–4.0 keV range and the thermal model fit (solid line) based on the 280 X-ray photons in the archival XMM-Newton field. Plots from Mullis et al. (2005) and H. Böhringer. 5.5. THESIS DATA OVERVIEW 75

Data NoFields Raw Data Reduced Data Survey Fraction XMM X-ray archive 546 280GB 600GB 100% NIR imaging 56 112GB 480GB 40% optical spectroscopy 1 (+35) 3 GB 8 GB 2%

Table 5.2: Thesis data overview. The data type, number of diﬀerent ﬁelds, and data amounts are indicated. The last column gives an estimate of the corresponding data fraction for the ongoing XDCP survey phase.

5.5 Thesis Data Overview

Before turning to the detailed discussion of the individual survey steps in the next three chapters, a data overview is provided in Tab. 5.2. One of the prime goals for the thesis is a systematic XMM-Newton archival survey for serendipitous distant cluster sources with subsequent follow-up observations following the survey steps 1–5 of Sect. 5.2. In total, the following chapters are based on more than 1 TB of processed data products. The XDCP survey area for the thesis project phase is defined by the XMM archival status as of November 2004 and consists of approximately 550 X-ray fields. The sky coverage is thus more than quadrupled compared to the pilot study and requires the processing of about 280 GB of raw X-ray data, which will be presented in the next Chap. 6. The second large amount of data included here has been acquired through several follow- up near-infrared imaging campaigns as discussed in Chap. 7. These imaging observations cover about 40% of all identified distant cluster candidates. The spectroscopic confirmation of the distant cluster candidates is currently gaining momentum but is still at an early stage with only few XDCP designated observations available. Achieving a spectroscopic survey completion will be the most challenging part and will realistically require another five years. The thesis work included the spectroscopic analysis of 35 low-redshift clusters for the NORAS 2 survey (see Sect. 4.2) which is not further discussed here. In order to maintain the balance for the current project status, Chap. 8 covering the spectroscopy is kept short and is focussed on the main aspects. 76 CHAPTER 5. THE XMM-NEWTON DISTANT CLUSTER PROJECT Chapter 6

X-ray Analysis of XMM-Newton Archival Data

In this chapter, we will discuss in detail the X-ray selection of distant cluster candidates. The first three sections will introduce the X-ray data, survey fields, and source detection procedures (survey step 1). Section four will review the source classification and candidate selection (survey step 2), and the final two sections will present test results and general diagnostic plots. The X-ray analysis constitutes the backbone of the XDCP survey and is crucial for the overall success of the project. The final goal of this X-ray part is the compilation of a well-selected sample of distant cluster candidates which should (i) exploit the maximum sensitivity achievable with XMM archive data, (ii) minimize the low-redshift cluster fraction, and (iii) include only an acceptable fraction of false positive sources while remaining highly complete for true high-redshift clusters. The assessment of the extended nature of X-ray sources is of prime importance for the cluster selection process. Estimations of the source extent and the X-ray flux are the main accessible observables at this point of the survey chain. In combination with a qualitative evaluation of the X-ray morphology, first clues to the dynamical state of the cluster sources can be obtained. After completion of survey steps 3–5, the X-ray source properties can be transformed to physical object quantities such as luminosity, temperature (for brighter sources), and a mass estimate.

6.1 XMM-Newton in a Nutshell

The XMM-Newton data archive constitutes the basis and starting point of the XDCP survey. A detailed understanding of the data properties, capabilities, and limitations is hence the pre-requisite for a proper analysis and for the survey application. In this introductory section, the main XMM-Newton characteristics related to distant cluster searches are summarized and evaluated.

77 78 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Figure 6.1: Left: The XMM-Newton observatory with its three X-ray telescopes. Right: Eﬀective area as function of energy. The total XMM-Newton eﬀective area for X-ray imaging consists of the sum of the two upper curves. Plots from Ehle et al. (2007).

XMM-Newton, the Xray Multi-Mirror Mission, was launched on 10 December 1999 as the second of ESAs four cornerstone missions of the Horizon 2000 program. With a length of more than 10 m and a weight of almost 4 tons, XMM-Newton is the largest scientific satellite ever built in Europe (see Fig. 6.1). The observatory was launched into a highly elliptical 48 hour orbit with a perigee of 6 000 km and an apogee1 of 115000km allowing about 130 ksec of uninterrupted science observations per revolution outside Earth’s radiation belts2 . The main scientific payload consists of three co-aligned X-ray mirror systems made up of 58 nested mirror shells each. This large number of Wolter type X-ray mirrors provide an unprecedented effective geometric3 area of 1550cm2 per module. The total gathering power is shared by five simultaneously operating focal plane X-ray instruments, two Reflection Grating Spectrometers (RGS) and three European Photon Imaging Cameras (EPIC). The XMM-Newton EPIC imaging instruments use two different kinds of X-ray CCD technologies. The two identical MOS (Metal Oxide Semiconductors) cameras use seven front-illuminated 600 600 pixel CCDs with a frame store region as data buffer. Each MOS camera shares a mirror× module with one of the RGS spectrometers resulting in a reduced effective area for imaging of 44%. The more sensitive PN camera,4 on the other hand, receives the full reflected light of one of the three X-ray telescopes. The twelve backside- illuminated CCD chips with a total of 376 384 pixels are fabricated from a single Silicon × 1An orbital manoeuvre in February 2003 has changed the orbital parameters to the current perigee and apogee values of 20000km and 101000km (Ehle, 2007). 2Science observations can typically be conducted at elevations above 46000km. 3The combined area, i.e. the sum of all thin shell-like regions from where (on-axis) photons are focussed onto the detector, is equivalent to the area of a circle with a diameter of 44.4 cm. 4PN is derived from the pn-junction of the silicon semiconductor technology used in the detectors. For easier readability, the capitalized form PN, instead of pn, is used throughout the thesis. 6.1. XMM-NEWTON IN A NUTSHELL 79

wafer. They have a higher quantum efficiency (QEPN > 90%) than the MOS detectors (QE 40–85%), and allow higher frame rates through the integration of readout nodes MOS ∼ for each individual pixel column. The PN chips lack frame store buffers, which in practice has the consequence of so-called out-of-time events (OoT). Since the pixels still register incoming X-ray events during the column readout phase, which lasts a few percent of the full integration cycle,5 bright sources will imprint a smeared event streak over the full pixel column (Y direction) due to falsely identified photon positions during readout. For survey applications, we are mainly interested in full-frame imaging mode observations of the instruments.6 In contrast to optical or NIR pixel arrays (see Chaps. 7 & 8), X-ray CCDs are operated in single photon counting mode enabling non-dispersive low resolution spectroscopy. Incident X-ray photons produce a number of recorded electrons inside the detector pixel via an internal photoelectric effect which scales with the X-ray energy. If the CCD readout is sufficiently fast to ensure at most a single X-ray photon interaction per pixel neighborhood7 and clocking cycle, the energy resolution is achieved by matching the measured pixel signal to a calibrated energy.8 Table 6.1 summarizes the most important XMM-Newton imaging mode characteristics. After the general overview, we will now discuss the main instrumental factors for distant galaxy cluster searches ordered with decreasing priority.

Effective Area: The effective area Ae is the prime instrument characteristic related to the achievable sensitivity per unit time. The effective instrumental photon gathering power is determined as the product of the geometric area of the mirror modules, their reflectivity, and the quantum efficiency of the detectors, shown in the right panel of Fig. 6.1 for all XMM-Newton camera systems. The solid black line illustrates Ae as a function of photon energy for the PN camera as the most sensitive imaging instrument. The two MOS9 detectors provide lower quantum efficiencies and are only exposed to 44% of the geometric mirror area resulting in the dotted blue line for the individual instrument and the solid blue line for the sum of both modules. The 2 total effective area for all three XMM imaging instruments adds up to Ae 2 500 cm at 1 keV. XMM-Newton achieves a photon gathering power which outperforms≃ the Chandra observatory sensitivity by about a factor of four, and roughly by a factor of ten in comparison to ROSAT (e.g. Romer et al., 2001). The grazing-incidence reflection efficiency of X-ray photons on the mirror shells is very sensitive to the incident angle with respect to the optical axis of the module. This

5The out-of-time fraction is 6.3% for full frame mode and 2.3% for the extended full frame mode. 6Various other timing, burst, or small window modes are available for different science applications. 7X-ray photons can deposit their energy over several connected pixels, a pixel cluster, which have to be isolated spatially and temporally for an accurate energy reconstruction. Typically, only single and double pixel events are considered for the science analysis. 8Pile-up events, i.e. several photons recorded in the same pixel, can be identified and discarded during the data processing. 9The PN camera is effectively a factor of 3.3 more sensitive than the individual MOS cameras at 1 keV. A factor of 2.3 is attributed to the different geometric mirror area, the remaining factor of 1.5 is due to the QE differences. 80 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

gives rise to the vignetting effect, i.e. the decreasing effective area with increasing off-axis angle from the optical axis, which is shown in the left panel of Fig. 6.2. At Θ 10.5 arcmin the photon gathering power has dropped to 50% of the central value≃ and continues to fall towards the outskirts of the FoV. An integration of the vignetting function over the detector area out to a given off-axis angle yields the ′ ′ average vignetting factor V¯vig( Θ) as V¯vig( 10 ) 0.72, V¯vig( 12 ) 0.65, and V¯ ( 15′) 0.55. ≤ ≤ ≃ ≤ ≃ vig ≤ ≃ Field-of-View: The total XMM-Newton field-of-view is about 30 arcmin in diameter, corresponding to 0.2 deg2. The relevant instrument characteristic for survey applications is the product∼ of the FoV solid angle and the effective area, the so-called grasp g Ae Ω, which determines the required time to cover a unit sky area to a specified depth.≡ · Taking the average vignetting factors into account, we obtain (at 1.5 keV) 2 2 2 2 g15 270 cm deg for a maximum off-axis angle of 15 arcmin, and g12 204 cm deg ≃ 2 2 ≃ and g10 157 cm deg when restricting to FoV to off-axis angles of 12 arcmin or 10 arcmin,≃ respectively. The motivation to constrain the field-of-view for a cluster survey arises from the PSF properties discussed below.

Spatial Resolution: The first two items determine the number of photons per source and unit time and the total sky area per observation. The most crucial point for any serendipitous cluster survey is the assessment of the source extent properties, which serve as primary selection criteria. Simply speaking, a source can be spatially resolved if its extent is larger than the instrumental point-spread-function (PSF). This implies that an accurate knowledge of the PSF shape at the source position is the pre-requisite for determining a spatial extent of the source. As was the case for the vignetting factor, the PSF varies strongly as a function of off-axis angle. Figure 6.3 illustrates the changing shape of the XMM point-spread-function in 3 arcmin steps from the center. The increasing asymmetry and complexity towards the FoV outskirts is obvious, and is the main reason for prevailing uncertainties in the off-axis PSF calibration even after more than seven years of XMM operations. The central PSF is well described by a King profile of the form

A P SF (r)= 2 α , (6.1) 1+ r rc with core radius rc, index α, and normalization A. All ﬁtting parameters are functions of oﬀ-axis angle and energy. The on-axis PSF core radii at 1.5 keV have been 10 ′′ ′′ ′′ determined as rc(MOS1) 4.7 , rc(MOS2) 4.5 , and rc(PN) 6.2 . The instru- ≃ ≃ ≃ ′′ ′′ mental XMM on-axis resolution limit is hence set at cluster core radii of rc 5 –6 , corresponding to a physical resolved cluster core scale at z > 1 (see Fig.3.1)≃ of lim rc 50 kpc. However, the on-axis PSF sets a lower ideal resolution limit, which is not∼ applicable outside the very center of the FoV. From the lower row of PSF images

10PSF calibration documentation EPIC-MCT-TN-012, 2002 by S. Ghizzardi. See http://xmm.vilspa. esa.es/external/xmm sw cal/calib/documentation/index.shtml. 6.1. XMM-NEWTON IN A NUTSHELL 81

Characteristic XMM Performance number of telescopes and imaging detectors 3 total eﬀective area at 1keV 2 500 cm2 ∼ ﬁeld-of-view 30′ diameter 0.2 deg2 spatial resolution (FWHM) 5′′–15≃′′ half-energy width (HEW) 14′′–20′′ pixel scale PN / MOS 4′′.1/1′′.1 per pixel energy resolution at 1keV 80 eV ∼ point source sensitivity in 10 ksec (all cameras) 5 1015 erg s−1cm−2 ∼ × Table 6.1: XMM-Newton imaging-mode characteristics.

in Fig. 6.3 it becomes obvious that the azimuthally symmetric King profile is not a good approximation to the increasing asymmetry of the core. Meaningful calibration data characterizing the off-axis PSF is sparse, and a formal application of the King profile fit even returns a decreasing core radius at increasing off-axis angles. A more practical illustration of the off-axis dependence on the resolution limit is given in the right panel of Fig. 6.2, which shows Gaussian full-width at half-maximum (FWHM) measurements of sources in three deep fields. The combined photon images of all three detectors were used, thus returning the properties of the average XMM PSF. The red line in Fig. 6.2 indicates the averaged results and is consistent with the expected off-axis trend of a FWHM broadening from about 8′′ in the center to roughly 14′′–15′′ in the detector periphery. The uncertainty in the off-axis PSF calibration is one of the main reasons for the detection of spurious extended sources and has to be adequately taken into account for the source classification.

Background Characteristics: The X-ray background is the second critical issue for the detection of extended sources, as uncertainties in the model background can also result in spurious extended sources. For low surface brightness objects such as clusters, the X-ray background is the prime limitation of the achievable sensitivity. Concerning the background properties, XMM-Newton is not in an ideal orbit (see Sect. 6.6.1) as observations are contaminated from frequent background flaring events and exhibit rather high quiescent levels. This has the consequence that cluster studies require significant efforts for the suppression and proper treatment of the X-ray background (see Sect. 6.3.2).

Energy Resolution: The XMM-Newton energy resolution in imaging mode of 80 eV ∼ at 1 keV is the least critical among the discussed items for distant cluster searches. However, it enables (i) the a posteriori definition of band-passes, (ii) the assessment of hardness ratios for all sources, and (iii) more detailed spectroscopic studies of confirmed clusters, e.g. for a temperature determination of sufficiently bright sources. 82 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Figure 6.2: XMM vignetting and PSF properties as function of off-axis angle. Left: Average vignetting function of all three XMM X-ray telescope systems, with the PN optical axis used as central reference point. The horizontal dashed lines indicate the 100% and 50% effective area levels, the latter is reached at an off-axis angle of about 10.5 arcmin. The sharp drop starting at 14.5 arcmin off-axis ∼ angle indicates the onset of the FoV edge of some detectors. Right: Gaussian FWHM of sources as determined in three deep fields. The measurements were performed in the combined image of all three detectors. The error bar length is proportional to the ratio of the FWHM in the X and Y direction, i.e. to the amount of PSF asymmetry. The red solid line illustrates the smoothed trend of the average FWHM between about 8′′ in the center and 15′′ at the FoV edge.

Figure 6.3: XMM PSF shape variations as a function of oﬀ-axis angle. Shown are the 2-D PSF images of the PN mirror module in oﬀ-axis angle steps of 3 arcmin as provided by the XMM calibration data base. The increasing asymmetry towards the FoV edge poses a bigger problem to the detection of extended sources than the moderate overall broadening of the PSF core. 6.2. X-RAY DATA ARCHIVE AND SURVEY DEFINITION 83

6.2 X-ray Data Archive and Survey Deﬁnition

All five X-ray instruments of the XMM-Newton observatory are operating simultaneously during targeted observations. Even for science programs that are not focussed on XMM’s imaging capabilities, any idle PN or MOS camera proceed with the data acquisition. All observations are collected in the online XMM data archive11, where they become publicly available after a proprietary period of typically twelve months. The XMM archive is hence constantly growing with now more than seven years of successful operations. In order to obtain a well controlled sample for the current project phase, the general XDCP survey field definition has been frozen at the archive status of 2 November 2004. After almost five years of in-orbit operations, the XMM data archive exhibited the following statistical properties12: 2960 fields with expired proprietary period were available in total with a total nominal • exposure time sum of Σ 72.3 Msec; ≃ 2016 of these observations had at least one camera operating in full frame imaging • mode; 1910 remained after rejecting sky regions ‘contaminated’ by the Large and Small • Magellanic Clouds and by the Andromeda galaxy M31; 1109 fields were located at sufficiently high galactic latitudes with b 20◦ and had • | | ≥ nominal exposure times of more than 10 ksec (see dotted black line in Fig. 6.4 for exposure time distribution); 1018 observations were not attributed to designated survey programs with sepa- • rate source identification programs, e.g. XMM-LSS, COSMOS, Chandra Deep Fields (black solid line in Fig. 6.4); 575 fields were VLT accessible at declinations DEC +20◦ (blue solid line) defining • ≤ the XDCP raw survey sample with the sky distribution shown in Fig. 6.5; 546 data sets (see field list in Tab. A.1) passed a first quicklook check and were • processed with the XDCP X-ray reduction pipeline (Σ 17.5 Msec); ≃ 469 fields were fully analyzed and are in principle suitable for distant cluster searches, • defining the XDCP core survey sample (Σ 15.2 Msec); ≃ 106 analyzed observations are located within the South Pole Telescope SZE survey • region at DEC< 30 deg and 21 h < RA < 7 h indicated by red symbols in Fig. 6.5 (Σ 3.3 Msec). − ≃

11The XMM-Newton data archive can be accessed through http://xmm.vilspa.esa.es/external/ xmm data acc/xsa/index.shtml. 12The software for processing XMM archive lists and for automated data download was originally written by Chris Mullis. 84 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Figure 6.4: Cumulative distribution of XMM-Newton archival ﬁelds suitable for a serendipitous survey as of status November 2004. The blue solid line shows the Southern sample (DEC< +20◦) of 575 XMM-Newton observations used for the analysis in this thesis. The black solid line depicts the full sample including the Northern sky, dashed lines include in addition observations of designated deep surveys (e.g. XMM-LSS, COSMOS, see Sect. 4.4).

Figure 6.5: Sky distribution of the 575 Southern XMM-Newton ﬁelds suitable for a survey (raw XDCP sample). The red squares indicate ﬁelds which will be covered by SZE observations with the South Pole Telescope. Square symbols are not to scale. 6.3. DEVELOPMENT OF AN X-RAY REDUCTION AND ANALYSIS PIPELINE 85

6.3 Development of an X-ray Reduction and Analysis Pipeline

In this section, the basic concepts of the X-ray data reduction and source detection processes are introduced. The handling of hundreds of XDCP survey field data sets required the development of a designated, distant cluster-optimized X-ray reduction pipeline. The driver and final goal of the X-ray pipeline development is the maximally achievable sensitivity for faint extended cluster sources in the archival data. This development is based upon the available tools of the XMM Science Analysis Software13 (SAS), the mission software package to reduce and analyze XMM data.14 In particular for the critical source detection procedure, the performance of the SAS tasks impose a ‘boundary condition’ for the survey outcome. The overall reduction and analysis approach aims at the compilation of a versatile raw survey masterlist, obtained by homogeneously applying the same procedure to all fields of the predefined XDCP sample. These raw results should comprise a super-set of the finalized survey sample in terms of number of analyzed fields, detector area, significance thresholds, and source characterization and should enable cross-checks and performance evaluations. The final XDCP science survey sample can then be defined a posteriori based on optimized and more restrictive cuts concerning (i) the flux limit, (ii) significance thresholds, (iii) usable detector area, and (iv) survey field selection. Since the XDCP is pushing the limits of what is achievable with XMM for distant cluster detection, this versatile approach promises the most flexibility for selection adjustments and the final survey characterizations once additional observational feedback is available.

6.3.1 Search strategy For faint object detections in any waveband, the crucial characteristics is always the signal- to-noise ratio (SNR), i.e. the source signal in a given aperture divided by the total noise in the same area. A first possible distant cluster optimization is hence related to the energy range of the applied detection bands. As discussed in Sect. 2.3, the observed thermal bremsstrahlung spectra of plasmas at temperature T (Equ. 2.2) are characterized by an exponential cut-off beyond energies of E k T at the high-energy end and a galactic cut ≃ B hydrogen absorption suppression at the low-energy end of Eabs <0.3 keV. While the latter spectral boundary is redshift-independent, the first one scales∼ as E (z) k T/(1+z). cut ≃ B A fiducial T 5 keV cluster at z 1 will hence exhibit the spectral break at about 2.5 keV, which∼ will shift towards lower∼ energies at even higher redshifts. These simple considerations imply that distant cluster searches should be focussed on the soft bands of the XMM sensitivity range. Scharf (2002) has determined the optimal XMM detection bands for yielding the highest SNR for galaxy clusters and groups with different temperatures and redshifts. Figure 6.6 summarizes these results for the MOS detectors (left panel), the PN instrument (center

13The SAS homepage can be found at http://xmm.vilspa.esa.es/sas. 14SAS tasks such as eboxdetect are set in typewriter font for easy identiﬁcation. 86 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

panel), and the SNR improvement for the PN (right panel). The XDCP prime target objects are z >1 clusters with temperatures 2 keV

(1) Single Band Scheme: The conceptually simplest search strategy for distant clusters is the use of a single detection band. As discussed, the 0.35-2.4 keV band is expected to optimize the target signal-to-noise ratio and the total number of source photons over a wide range of the expected source characteristics. This band-pass is used as the primary XDCP detection scheme and almost coincides with the ROSAT band (0.1– 2.4 keV). A single deﬁned energy band will simplify the survey calibration and the determination of the selection function through extensive simulations (see Sect. 9.1.3).

(2) Spectral Matched Filter Scheme (SMF): The first detection scheme does not allow the determination of spectral hardness ratio properties and additionally makes use of justified but fairly strong priors on the expected spectral source properties. It is conceivable that very hot systems at T >10 keV exist in the distant Universe, e.g. as the result of a recent major merger event.∼ In order to broaden the energy baseline and hence increase the sensitivity for hotter objects, the second scheme covers the full 0.3–7.5 keV range, but assigns different weights to parts of the spectral range, as shown by the blue line in Fig. 6.7. By enhancing the softer energy regions, the shape of the weight function roughly follows the (expected) continuum form of hot distant clusters (see lower right panel of Fig. 5.2) and is consequently termed spectral matched filter scheme (SMF). The effective weight function is achieved in practice by using five (partially) overlapping15 bands (0.3–0.5, 0.35-2.4, 0.5–2.0, 2.0–4.5, 0.5– 7.5 keV). The SMF scheme is to be considered as an experimental approach, which has yielded promising performance results in early comparison tests. The empirically

15For the source characterization and the associated likelihoods, this implementation is statistically equivalent to the use of a single broad band (0.3–7.5 keV) with diﬀerent assigned weights over the spectral range. 6.3. DEVELOPMENT OF AN X-RAY REDUCTION AND ANALYSIS PIPELINE 87

determined likelihood oﬀset factors for the signiﬁcance evaluation and the comparison to the other schemes will be discussed in Sect. 6.6.2.

(3) Standard Scheme (STD): The third applied scheme makes use of the combination of three XMM standard bands in the range 0.3–0.5 keV, 0.5–2.0 keV, and 2.0–4.5 keV, i.e. it provides continuous coverage from 0.3–4.5 keV. This approach is widely used and can hence be considered as the standard scheme.

Figure 6.7 summarizes all three detection schemes and their energy coverage. All methods oﬀer some advantages and complementarity. The ﬁrst detection scheme is attractive for its simplicity, the second for its additional sensitivity to ‘exotic systems’, and the third for its standardized source parameters. Three main reasons motivated the application of a complementary multi-scheme detection method:

achieving the best possible survey completeness of detected extended sources; • enabling a cross-comparison of methods to test for systematic selection eﬀects; • increasing the reliability of the source classiﬁcation close to the detection threshold • by evaluating the stability of important source parameter solutions.

6.3.2 X-ray data reduction The general outline of the developed XDCP X-ray reduction pipeline is illustrated in Fig. 6.8. Raw XMM-Newton data for each observation are organized in a so-called Ob- servation Data File (ODF), a directory with a collection of 200 files containing the ∼ uncalibrated data of all instruments, satellite attitude files, and calibration information. ODFs have unique identifications numbers (OBSID) and can be considered as the basic starting data set from which all science products can be derived. An arbitrarily long list of ODFs serves as the input for the X-ray pipeline and is passed on to the top-level script, which organizes the ODF handling and the calls of the individual reduction modules. In practice, 1–3 modules are applied to typically 100 data sets at a time, followed by quality and completion checks of the intermediate data products. All data sets have been homogenously16 reduced with the XMM Science Analysis Software version SAS 6.5. The full XDCP pipeline can be subdivided into three main parts: (i) The actual X- ray data reduction modules, which create calibrated science images from the raw satellite data, (ii) the source detection and characterization procedures resulting in lists of extended sources, and (iii) the interactive individual detailed source analysis and classification using X-ray-optical image overlays. In this sub-section, we will have a closer look at the data reduction modules17 of the first part.

16An early reduction run of 120 XMM observations using SAS6.1 has been repeated with SAS6.5, released in August 2005. 17The individual X-ray reduction modules are adapted and extended versions of codes originally written by Hans B¨ohringer. 88 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Figure 6.6: Optimized X-ray source detection bands for galaxy clusters and groups as derived by Scharf (2002). Left: Upper (black solid lines) and lower (black dotted lines) detection band-pass versus object redshift for the XMM MOS detectors for systems with diﬀerent temperatures. From top to bottom the assumed cluster/group temperature is 6, 4, 2, 1, 0.5, 0.2 keV. The blue solid lines illustrate the XDCP detection band with an energy range of 0.35–2.4 keV, which was chosen to yield an optimized high-redshift detection performance over a wide range of temperatures. Center: Same plot for the XMM PN detector. Right: Expected PN signal-to-noise ratio of the standard 0.5–2.0 keV band divided by the SNR of the optimized bands (black lines) in the center panel. The SNR gain is most notable for lower temperature systems at T <2 keV. Plots adapted from Scharf (2002). ∼

Figure 6.7: Detection schemes. The relative photon weight factor is plotted against the energy for the unweighted single band scheme (red line), the three-band standard scheme (green), and the spectral matched filter scheme (blue) featuring different photon weights across the spectral range. Small offsets are applied for easier distinction of band boundaries. 6.3. DEVELOPMENT OF AN X-RAY REDUCTION AND ANALYSIS PIPELINE 89

XDCP X-ray Pipeline

ODF 3 ODF 2 ODF 1

X-ray Data Reduction

Data Preparation CAL organize satellite data & calibration ﬁles

Data Processing Calibrated run processing chains for all detectors Event Files

Clean Statistics Iterated Flare Cleaning 1) hard-band ﬂare removal Cleaned 2) soft-band ﬂare removal Event Files

Image Creation for diﬀerent bands and all detectors

Combination & Smoothing Smoothed PN+MOSs combined & smoothed images X-ray Images

Source Detection

Wavelet Detection Wavelet ewavelet Detected Sources Exclusion Mask mask out contaminating sources

CAL Exposure+Background Maps for all bands

Box Detection & ML Fitting eboxdetect+emldetect for diﬀerent methods

Source Analysis+Extraction Extended display sources & extract source list Source Masterlist

X-ray-Optical Overlays

X-ray Contours for combined and single detectors DSS X-ray- Create Overlays Optical NED DSS+X-ray+NED source zoom Overlays

Figure 6.8: X-ray reduction and analysis pipeline ﬂow chart. 90 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

XMM data setup The first task is to transform the XMM sets of raw satellite data, each ODF containing typically 500 MB of data, into a suitable format for further processing. This is achieved with few straightforward∼ SAS procedures which keep the complexity of the underlying raw data hidden from the user. The task cifbuild matches the input data with the appropriate calibration files, which are accessible through the so-called Calibration Access Layer (CAL). odfingest extracts and organizes the satellite housekeeping data necessary to accurately reconstruct the observations. The instrument specific processing chains epchain for the PN and emchain for the MOS cameras generate calibrated photon event lists for the complete observation with reconstructed sky positions, energies, arrival times, and event characteristics for each individual X-ray photon. In Sect. 6.2, one of the survey field selection criteria was a minimal nominal exposure time of 10 ksec, which is the period XMM was pointing at the target location. However, the nominal on-target times are not equivalent to the effective science-usable clean exposure times, which can be significantly lower for two main reasons. (i) Instrument calibration overheads at the beginning of an observation require about 0.5–1.5 h for the PN, and 10 min for the MOS instruments in order to compute the optical loading of the CCD chips, i.e. the signal fraction attributed to optical light rather than X-ray photons. The more sensitive PN camera hence achieves on average about 1 h less on-target time than the MOS instruments. (ii) In addition to the constant hardware overheads, the more or less unpredictable space weather can contaminate observations resulting in extended periods of lost time. The space weather effects and the different XMM background components are discussed in the following.

X-ray backgrounds and flare cleaning The XMM background is the prime constraint for the achievable sensitivity limit of low surface brightness extended sources. For its importance, the main background compo- nents18 are shortly introduced followed by a discussion of the implemented flare cleaning procedure. Cosmic X-ray Background (CXB): The cosmic X-ray background can be separated into two main contributions. (i) The hard component of the CXB dominates at energies >1 keV and is mostly attributed to the integrated light of faint, unresolved extragalactic∼ X-ray sources, predominantly Active Galactic Nuclei (AGN). Extremely deep 1 Msec pencil-beam observations with Chandra have resolved the largest fraction of the hard CXB into individual point sources (e.g. Mushotzky et al., 2000; Giacconi et al., 2001). The total CXB flux in the 2–10 keV hard band reaches an approximate surface brightness of 5 10−15erg s−1cm−2arcmin−2. (ii) The soft component of the × CXB, dominating below 1 keV, can be attributed to thermal diffuse emission of galactic origin. It originates from plasma of the Local Hot Bubble, the Galactic Disk, and 18All XMM background components are summarized by A. Read at http://www.star.le.ac. uk/%7Eamr30/BG/BGTable.html. 6.3. DEVELOPMENT OF AN X-RAY REDUCTION AND ANALYSIS PIPELINE 91

the Galactic Halo and is consistent with thermal emission of about T 0.1–0.2 keV. The CXB is constant in time, but the soft component varies spatially across∼ the sky.

Flaring Particle Background: This background component was severely underestimated before the launch of XMM and is responsible for most of the lost science time. The ﬂaring background has strong rapid variability and is attributed to soft protons of solar origin with energies of a few 100 keV. These protons, which seem to be organized in ‘clouds’ populating the Earth’s magnetosphere, are funnelled by the X-ray mirrors towards the detectors, where they typically deposit several keV of their energy in the CCD pixels. The occurrence of the soft proton ﬂares is unpredictable but is related to the solar activity and the satellite position with respect to the magnetosphere. Flare intensities can reach 1 000 times the quiescent background level implying that contaminated time intervals have to be carefully cleaned and removed from the science data. Since the soft protons pass the mirror system, they imprint a (partially) vignetted background component onto the detectors.

Quiescent Particle Induced Background: This second particle induced component dominates the quiescent background level. It is attributed to high-energy cosmic ray events of some 100 MeV, where a charged particle interacts with the detector material and the surrounding structures to produce fluorescence emission. This more stable internal background component typically varies only on the 15% level during observations, but can be an order of magnitude enhanced in high± radiation periods. The cosmic rays induce a background component originating from the detector material itself, i.e. the quiescent particle background is not vignetted, but reflects the spatial structure of the surrounding detector material. The most prominent instrumental fluorescence lines are due to the Al-Kα transition at 1.5 keV for all instruments and additional Si-Kα emission at 1.7 keV for the MOS detectors. Another intense line complex of several elements (e.g. Cu, Ni, Zn) arises around 8 keV for the PN and is hence beyond the considered bands for the survey.

Instrumental Noise: The last considered background contribution is electronic detector noise. This component is negligible in the harder bands but becomes signiﬁcant at low energies of <0.3 keV. The low energy band limit for the survey of 0.3 keV is thus motivated by the∼ onset of increased instrumental noise in conjunction with enhanced galactic hydrogen absorption.

The top left panel of Fig. 6.9 shows a typical 100 sec-binned light curve in the PN 12- 14 keV hard band of a 47 ksec pointing. The soft proton flare during the central part of the observation is obvious and can be particularly well identified in the hardest band due to the flat nature of the flare spectrum. The pipeline flare removal procedure of the first cleaning stage automatically identifies the stable quiescent level based on a Gaussian peak-fit in the count-level histogram of the time series (center left panel of Fig. 6.9). All observation periods with background levels less than three standard deviations above the determined quiescent level are accepted as good time intervals (GTI) for the initial hard band cleaning. 92 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Figure 6.9: X-ray flare cleaning procedure. Top row left: 100 sec-binned PN light curve of the complete field in the hard X-ray band at 12–14 keV. The flared periods are easily visible and are removed during the first cleaning stage. Right: 10 sec-binned PN ‘soft’ band light curve at 0.3– 10keV. The missed rising flanks close to the onset of flare periods and short flares are identified during the second cleaning stage. Center row: Automatic good time interval analysis based on the identification of the quiescent level and specified κσ-clipping cut-levels for cleaning stage one (left) and stage two (right). Bottom row: Final light curve of the identified good time intervals for the PN (left) and MOS (right) detectors. 6.3. DEVELOPMENT OF AN X-RAY REDUCTION AND ANALYSIS PIPELINE 93

This first cleaning stage efficiently removes the largest fraction of flare-contaminated data. However, at lower energies the flares can exhibit a different time behavior than the main hard component, which is currently attributed to lower energy particles at the edges of the encountered proton ‘clouds’. The upper right panel of Fig. 6.9 displays the 10 sec- binned light curve in the full 0.3–10 keV soft X-ray band after the initial flare removal. The residual flare peaks are identified during the second soft-band cleaning stage19 and are removed from the remaining good time intervals via 5-σ clipping of the soft energy histogram (center right panel of Fig. 6.9). The lower panels of Fig. 6.9 show the resulting light curves (100 sec-binned) after the iterated flare cleaning procedure for the PN instrument (left) and one of the MOS cameras (right). The contaminated time intervals are completely removed from the final cleaned photon event lists resulting in an effective science-usable exposure time of 29.6 ksec or 63% of the nominal time for this field. The flare removal procedure is applied to all instruments independently since the higher low energy sensitivity of the PN camera results in a different temporal response of the system. For about 80% of all XMM fields, the automatic double flare cleaning procedures yields good and stable results in identifying the good time intervals. If more than half of the observation is contaminated, the automatic determination of the quiescent background level might fail and require a manual definition of the GTI cuts.

X-ray images With the cleaned photon event lists at hand, we can now create X-ray images for all subsequent analysis tasks. The energy information of the X-ray photon events allows the arbitrary definition of customized band-passes at this point. As discussed in Sect. 6.3.1, five different energy bands are considered for the XDCP survey. Photons of the 0.35–2.4 keV energy range are selected for the optimized single detection band, the ranges 0.3–0.5 keV, 0.5–2.0 keV, and 2.0–4.5 keV for the standard band method, and an additional 0.5–7.5 keV broad band for the spectral matched filter scheme, which uses all five bands. The X-ray images are reconstructed with 4 arcsec pixels from the cleaned photon event lists by applying the appropriate energy cuts and conservatively selecting only well calibrated events as characterized by event quality flags. The absolute astrometry, i.e. the sky coordinate information, is typically accurately determined to within approximately one arcsecond. Images are produced for each instrument individually resulting in a total of 15 band-detector combinations. At this stage, the correction for out-of-time events of the PN detector is performed. Based on the observed count rate in each pixel, the smeared streaks of photons with mismatched Y position information can be statistically reconstructed in an OoT image20. This OoT frame is created for each band and subtracted from the corresponding raw PN data to yield a first order correction of the out-of-time image artifacts. Although the fraction of smeared photons is only 2.3% or 6.3% (depending on the imaging mode), the trails of very bright central target objects can mimic apparently extended sources and should therefore be corrected. 19The double flare cleaning procedure was originally proposed and applied by Pratt and Arnaud (2003). 20A reconstructed OoT event list can be obtained with the task epchain. 94 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Observations that have been interrupted, e.g. due to technical problems or strong flares, are often split up into several shorter data blocks per instrument. The separate event lists are individually reduced, cleaned, and transformed into X-ray images. These shorter exposures are co-added for each band-pass and instrument at this point in order to restore the full on-target integration time of the observation. As a last step of the data reduction part, images of the same band are combined for the three detectors to yield the full XMM imaging information in the specified energy range. In order to enhance the contrast of weak sources for visual inspection, the combined images are additionally smoothed with a 4 arcsec Gaussian filter21. Figure 6.10 displays the raw (center left panel) and smoothed (upper left panel) combined 0.35–2.4 keV images of one of the deep XMM survey fields. The first part of the XDCP pipeline is now finished and we can proceed with the source detection module.

6.3.3 Detection of extended X-ray sources The source detection procedure22 as shown in Fig. 6.8 requires several additional preparatory steps and intermediate data products. Figure 6.10 illustrates the X-ray image of the current reduction status (top left panel), the display of the final X-ray detection results (upper right panel), and a selection of intermediate data products in the numbered center and bottom panels. The first important calibration frames besides the raw photon count images (panel 1) are the exposure maps (panel 2), which correct for the varying sensitivity across the XMM field-of-view, similar to flatfields in optical and NIR imaging (see Sect. 7.3.2). Dividing a raw photon count image by its associated exposure map results in a sensitivity-corrected count rate image with units [counts s−1], normalized to the central, on-axis values. These calibrated count rates differ from the· final physical flux units in each pixel only by a multiplicative energy-dependent constant, the (on-axis) count-rate-to-flux conversion factor. In effect, the exposure maps contain the effective, local integration times associated with each detector pixel, scaled to the on-axis exposure. The radial shape of the exposure maps is mainly due to the vignetting function23 (see Fig. 6.2), which accounts for the decreasing off-axis sensitivity. The task eexpmap also includes the calibration information on the energy dependent detector quantum efficiency, chip gaps, dead columns, the transmission of the optical blocking filters, and the detector field-of-view. Panel 2 shows the combined exposure map of all three instruments with dark regions indicating the most sensitive parts and lighter colors the areas with a reduced effective exposure time. The position of the optical axis, also named boresight, is indicated by the green cross. The circles illustrate the 12′ (green) and 14.5′ off-axis radius, with the latter one roughly coinciding with the FoV boundaries of the two MOS cameras.

21 The ﬁltering should average over several pixels to reduce the noise (pixel scale: 4′′/pixel), but should not exceed the PSF FWHM to allow a distinction between point sources and extended objects. 22The box detection procedure follows in part the scheme of Georg Lamer. 23As an alternative method, vignetting eﬀects can be accounted for by applying a weighting factor for each pixel. 6.3. DEVELOPMENT OF AN X-RAY REDUCTION AND ANALYSIS PIPELINE 95

Figure 6.10: X-ray source detection products. Top row left: Smoothed X-ray image in the 0.35–2.4 keV detection band of the field with the distant cluster RDCS J1252-29 at z = 1.24 in the center. Right: Source detection results in the same field with point sources marked in green, extended sources indicated by larger circles, and the bright source in the upper right corner excluded from the detection area. Center and bottom rows: Labelled, intermediate image products of the source detection procedure. The green circle in the exposure map marks the 12 arcmin off-axis radius, blue indicates the 14.5 arcmin circle coinciding with the FoV edge of the MOS detectors. 96 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

The next step is the generation of detection masks with emask, as shown in panel 3 for the PN detector. The masks define the detector regions over which the source detection is performed (black areas), excluding chip gaps and dead columns (white). In addition, manually defined regions with bright contaminating sources are also cut out and excluded from the detection area. In the example case in Fig. 6.10, the bright off-axis source in the upper right corner has been removed (manually), but more typically the central target objects are cut out to ease serendipitous detections. Excluding bright and large area sources helps in two ways: (i) the occupied region does not contribute to the effective area of the survey and is this way automatically accounted for, and (ii) the background model is improved. The crucial background maps for each detector and band are determined in the final preparatory step. The following procedure is applied. (i) A first sliding box detection run is performed with eboxdetect in local mode, i.e. the background is approximated locally around the detection cell. This yields a preliminary source list which is used to identify the positions of all sources and excise them from the field. (ii) The resulting ‘cheese image’ only contains X-ray background contributions and is used as input for the background modelling process with esplinemap.24 The task performs a two-component model fit for the linear combination of a spatially constant background contribution (quiescent particle induced background and instrumental noise) and a vignetted component (CXB and residuals of the soft proton particle background). The final background model result is shown in panel 4 of Fig. 6.10. The applied reconstruction method is physically motivated by the different contributing background components and has the main advantage of a smooth radial profile, but the disadvantage that the global fit over the 30′ FoV does not account for local variations. In any case, this model is preferable to the alternative spline fitting method, which accounts for local variations, but is also prone to producing numerous spurious extended sources as a consequence of local minima of the fitted splines.

Sliding box detection and maximum likelihood fitting The actual source detection is performed in two steps. (i) A list of source candidates is obtained with the sliding box method, followed by (ii) the subsequent detailed analysis and source characterization via maximum likelihood (ML) fitting. In order to achieve the best sensitivity for the candidate source list, the task eboxdetect is run in map mode, i.e. the improved reconstructed global background maps are used as input rather than the local approximation. In addition to the photon images and background maps, the exposure maps and the detection masks are supplied to the task. eboxdetect then takes a 5 5 pixel (20′′ 20′′) detection cell and slides it over the input images. At any image point,× the PSF-weighted,× integrated cell counts are compared to the expected background level and the likelihood for a random background fluctuation of this level is calculated. The sensitivity to extended sources is achieved, by iterating the procedure three times in conjunction with doubling the detection cell size to 10 10, 20 20, × × 24The task esplinemap can be used for spline fits or two-component model fits. 6.3. DEVELOPMENT OF AN X-RAY REDUCTION AND ANALYSIS PIPELINE 97 and 40 40 pixel boxes. All candidate sources above a specified combined likelihood are saved to× an input file for the detailed source characterization. The significance for the detection or the extent of an X-ray source is usually expressed in terms of the likelihood L= ln p (Cruddace et al., 1988), where p is the probability − Pois Pois of a Poissonian random background fluctuation of counts in the detection cell, which would result in at least the number of observed counts. The L likelihood values are linear, i.e. the total likelihood of a source determined from the full band width is equivalent to the sum of the Li values of the individual input bands. The probability of a source to be real preal and not originate from a statistical fluctuation is hence given by preal =1 pPois = 1 exp( L). For a selection of likelihood values L the following statistical probabilities− for − − a real physical source are obtained: preal(L =1) 0.632, preal(L =2) 0.865, preal(L =3) 0.950, p (L=4) 0.982, p (L=5) 0.993, p ≃(L=6) 0.998, p ≃(L=8) 0.9997, and≃ real ≃ real ≃ real ≃ real ≃ preal(L=10) 0.99995. The likelihoods are roughly related to positive outliers of Gaussian statistics with≃ n σ significance via n L/2, i.e. L = 4 corresponds approximately to 2-σ significance and L=6 to a 3-σ detection.∼ The task emldetect performs the final characterization of the candidate sources using the same input frames and the additional candidate source list of eboxdetect. A maximum likelihood PSF fit in the photon images is applied simultaneously for all input bands. Each candidate source is fit with free parameters for the source location (RA & DEC), the source extent, and the source count rate in each band. Before a source is classified as extended, emldetect attempts to fit two overlapping point sources to the photon distribution. Only if a single extended source fit yields a significant likelihood improvement in comparison to the 2-PSF fit of a double point source, the extended nature of the source is established. emldetect characterizes extended sources based on a King profile fit (Equ. 2.6) with a fixed standard beta value of β =2/3 and the core radius Θc (or equivalently rc) as free fit parameter. The observed photon distribution is to be compared to the King profile of the extended source convolved with the two-dimensional PSF shape at the specific detector location, which is a function of off-axis angle (Fig. 6.3), azimuthal angle (PSF rotation), and energy. If the maximum likelihood for the source extent, determined simultaneously for all input bands, exceeds a specified minimum threshold value the measured parameters for the extended source are saved to the final emldetect source list. In the case that the extent likelihood falls below the threshold, the corresponding point source parameters are determined and saved. For the XDCP search strategy of Sect. 6.3.1, the emldetect maximum likelihood fitting procedure is performed once for each of the three different detection schemes. The minimum detection likelihood for a source is fixed at L =6 ( 3 σ), whereas the extent likelihood is ∼ successively lowered for the different schemes. For the single band scheme with highest XDCP priority the minimum extent significance is set to 2.5 σ (L =5), for the spectral matched filter scheme25 to 2 σ (L 4), and for the standard∼ scheme to 1.5 σ (L =3). ∼ ≃ ∼

25 This is an eﬀective value. The corresponding emldetect parameters is set to Lmin = 10 but the enhanced weight of the central spectral region enters the total likelihood accordingly through the sum of overlapping bands. 98 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

The decreasing extent threshold is to ensure that the majority of sources are detected as a resolved objects for several detection schemes in order to allow a parameter cross- comparison. The aim for the raw extended source list as input for the screening procedure is a minimum extent significance of 2–2.5 σ as will be discussed in Sect. 6.4. Panel 5 of Fig. 6.10 displays the final reconstructed source image for the PN detector. Note the increasing tangential elongation of the PSF as a function of off-axis angle. De- tected sources are also marked in the smoothed combined photon image in the upper right panel, where green circles symbolize point source detections and additional blue or red circles identify the extended sources. The last panel 6 of Fig. 6.10 shows the PN sensitivity map of the observation, where light colors indicate the most sensitive central areas. The task esensmap approximates the local sensitivity limits for point source detections of a specified significance based on the exposure maps, the background maps, and the detection masks using Poissonian count statistics. Similar maps for the extended source sensitivity have to be calculated for the final XDCP survey assessment. The procedure for this more complex but crucial task is briefly discussed in Sect. 9.1.3.

Wavelet detection Besides the traditional sliding box method, wavelet techniques have become a popular alternative approach for the detection of extended X-ray sources (e.g. Rosati et al., 1995; Vikhlinin et al., 1995; Herranz et al., 2002). For the XDCP source detection procedure, the SAS task ewavelet26 is used as a complementary source detection method to provide additional extent information, in particular for the evaluation of faint sources. The basic concepts of the wavelet technique are briefly summarized. Wavelet functions allow a multi-scale analysis of images and enable a significant enhancement of the contrast of sources of different sizes against a non-uniform background. Wavelets can be considered as an extension of Fourier transformations with the difference that they are (i) localized in space, and have (ii) a zero normalization. Wavelets additionally satisfy (iii) translational invariance and (iv) scaling properties according to Wb,σ(x)=(1/σ)W ((x b)/σ), where W (x) is the wavelet function, b the translation, and σ the scaling parameter.− The most popular wavelet function for X-ray source detection is the so-called Mexican hat (MH) wavelet, which is related to the second derivative of a Gaussian function and can be written in its two dimensional form as

2 2 r r − r 2 a2 gMH = 2 2 e , (6.2) a − a ! with the wavelet scale a and r2 =x2 +y2. The MH has a Gaussian core with positive values out to radius r =√2 a and negative values outside.

26See the ewavelet manual at http://xmm.vilspa.esa.es/sas/7.1.0/doc/ewavelet/index.html for more information. 6.3. DEVELOPMENT OF AN X-RAY REDUCTION AND ANALYSIS PIPELINE 99

Figure 6.11: Wavelet source detection products of the same ﬁeld as in Fig. 6.10. Left: Wavelet- reconstructed background map for the PN detector. Center: Reconstructed PN source map. Right: Wavelet detection results for the PN (green) and the MOS (blue and pink) detectors. This example illustrates for the PN camera how a poorly determined background leads to numerous spurious extended sources in the center and right panel. The SAS ewavelet detection results are hence only used as complementary information for the source evaluation.

The basic principle of the ewavelet detection method works as follows: (i) The X-ray image is convolved with the Mexican hat wavelet (analogous to Equ. 3.33). In effect, the MH acts like a sliding cell, where the positive part resembles the source cell of scale a, and the negative part the background area. The normalization ensures that the wavelet transform of a homogeneous background is zero. Local maxima in the convolved image correspond to source candidates of wavelet scale a. (ii) If the significance of the local maximum in the wavelet transform is above a specified threshold, the source location and peak height at scale a is saved. (iii) The convolution procedure is repeated for various wavelength scales a, starting with the PSF size, and successively increasing the scale by factors of √2 up to a given fraction of the image size. (iv) The final scale of the identified objects corresponds to the highest peak in all wavelet transforms at the source position. The wavelet technique does not require a supplied external background, but rather reconstructs it from the data. The ewavelet determined PN background for the same XMM field as before is shown in the left panel of Fig. 6.11. In contrast to the box detection case, the wavelet reconstructed background traces local variations in addition to the off- axis dependence. The reconstructed PN source map in the central panel displays the effect of a poorly determined local background, in this case probably induced by the bright source, which cannot be excised for the ewavelet task. Numerous spurious extended sources are detected at the positions of local minima (lighter regions in left panel) in the background. The right panel illustrates the wavelet detection results for all detectors with sources and their scales overlaid on the smoothed photon image. Since the ewavelet source characterization is known to be less reliable than the maximum likelihood fitting method, the results of the wavelet detection, as visualized in the latter image, are used as an additional qualitative extent indicator for the cross-comparison with the sliding box detected extended sources. 100 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

X-ray source characterization As the final step of the source detection procedure, the extended sources for each XMM field are extracted from the final emldetect source lists associated with the different detection schemes. These field lists are then merged for all XMM observations resulting in a global extended source masterlist, which is the starting point for the screening procedure of Sect. 6.4. Listed in decreasing priority, the relevant emldetect parameters for the characterization of extended sources are (i) the maximum likelihood of the source extent (EXT ML), (ii) the detection maximum likelihood (DET ML), (iii) the source extent, i.e. the determined core radius (rc), (iv) the total number of detected counts per band, (v) the derived flux (fX), (vi) the off-axis angle of the source (Θ), (vii) the hardness ratios for the multi-band schemes (HR), and (viii) the total effective exposure at the source location. emldetect measures the total count rate of a source in a given energy band. Physical X-ray fluxes are obtained from the count rate values by multiplying with the Energy Con- version Factors (ECF), which depend on the instrument response in the considered energy band, the optical blocking filter, the initial photon selection cuts, the galactic hydrogen column density, and the spectral source properties. For galaxy cluster sources, this implies that the true ECFs are a function of cluster temperature, ICM metallicity, and redshift, which are a priori not known and additionally differ from object to object. Final accurate flux measurements for individual clusters hence require the full source characterization, which is generally not available prior to the redshift confirmation (i.e. survey step 5). For the global flux characterization of the source population in early survey stages, fiducial source fluxes are obtained by using standard, fixed Energy Conversion Factors.27 These ECFs are similar to the values for a fiducial distant cluster (z 1, T 5 keV) and have ∼ ∼ been confirmed for several sources to typically yield results with 10–20% accuracy in the standard 0.5–2.0 keV band. As the final point of the X-ray data discussion, we consider the main reasons for spurious detections of extended sources using the sliding box detection and maximum likelihood fitting method. Two types of spurious extended sources are to be distinguished: (i) spurious detections (SD), i.e. the X-ray source is not real, and (ii) spurious extents (SE) implying that the actual source is point-like or a blend of several point sources. It should be emphasized, that the determined detection and extent likelihoods are purely of statistical nature and cannot account for systematic errors of the supplied calibration data. In decreasing order of occurrence, the dominant reasons for false positive detections beyond the statistical expectations are:

1. secondary detections in wings of large extended sources (SD);

2. PSF residuals due to uncertainties in the oﬀ-axis calibration (SE+SD);

27Following the XMM-Newton calibration document SSC-LUX-TN-0059 (2003) at http://isar.star. le.ac.uk/pubdocs/docindex.shtml. A power law spectrum with a photon index of 1.7 and an absorbing 20 2 column of 3 10 cm− is assumed. × 6.3. DEVELOPMENT OF AN X-RAY REDUCTION AND ANALYSIS PIPELINE 101

3. detections at detector or chip boundaries (SE+SD); 4. blends of three or more point sources (SE); 5. background residuals due to a poor local determination (SD+SE); 6. out-of-time trail residuals (SD); 7. optical loading residuals caused by bright optical sources (SD).

The brackets indicate, wether the artifacts result mostly in spurious extents (SE), spurious detections (SD), or both (SE+SD). The upper right panel of Fig. 6.10 illustrates as example several spurious sources in the wings of the excised bright object in the upper right corner. Valtchanov, Pierre & Gastaud (2001) have benchmarked half a dozen techniques for the detection of extended X-ray sources in XMM-Newton data. The main conclusion is that there is no existing single detection technique without any drawbacks. Concerning the performance of eboxdetect in combination with emldetect, they confirm a very good detection rate and decent parameter reconstruction at the expense of numerous spurious sources, in particular related to the splitting of larger objects. ewavelet also shows a good sensitivity, but exhibits poor photometric parameter determinations. In any case, the detection of faint extended X-ray sources is a very challenging task. The available techniques and methods are still evolving, in particular for their application to ongoing missions such as XMM-Newton. This section has provided an overview of the principal methods and has discussed some of the complexities and issues arising therefrom. At this point, it should be emphasized again that one of the main objectives of the XDCP approach is to push the X-ray source detection to the very limit and fully exploit the achievable XMM sensitivity. The inevitable price for maintaining a high completeness level of real cluster sources at very faint flux levels is an increased contamination fraction at early stages of the survey confirmation process (steps 1–4), i.e. before the observationally expensive spectroscopy (see Chap. 8). The largest fraction of detected spurious extended sources can be removed during the screening procedure described in Sect. 6.4 (step 2); the remaining non-cluster sources will be identified during the follow-up imaging stage (steps 3 & 4) discussed in Chap. 7. For the challenging aim of finding and confirming the most distant X-ray clusters at z>1, using less than half of the (relatively inexpensive) imaging time for the identification of the remaining false positive sources seems to be an adequate and tolerable observational price to pay.

6.3.4 X-ray-optical overlays As a preparatory step for the source screening procedure of Sect. 6.4, the ﬁnal part of the XDCP pipeline handles the generation of X-ray-optical overlays for the individual detected extended sources. These overlays provide essential information for the ﬁnal assessment of (i) the X-ray quality of the source, and (ii) possible optical counterparts. 102 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Figure 6.12: X-ray-optical overlays for the source classification procedure as complementary information to the emldetect source parameters. Top left: Unfiltered 5′ 5′ photon image in the × 0.35–2.4 keV band with the smoothed contours overlaid and the core radius indicated by the blue circle. Top right: Individual detector contours for the PN (green) and the MOS instruments (blue and pink) as a means for cross-checking instrumental artifacts. Bottom row left: DSS-red image (2.5′ 2.5′) with smoothed X-ray contours and NED object information overlaid. The source is iden- × tified as the known cluster RDCS J1252-29 at z = 1.24 (red diamond), green circles indicate NED sources without redshift information. Right: Same material overlaid on the DSS NIR image.

Combining data of several different wavebands into a single image can be achieved through the use of contours, which are overlaid onto a second image. The first step for the overlay production is hence the generation of X-ray contours, which are obtained from the smoothed single detection band images at 0.35–2.4 keV. The contours are computed for logarithmically-spaced, pre-defined flux levels for the combined image and the three detector frames individually. For the assessment of the X-ray quality of the source, the combined contours are overlaid onto the raw X-ray photon image with a FoV size of 5′ 5′ centered on the source. In × addition the determined King profile core radius of the extended source is overplotted. Panel 1 of Fig. 6.12 illustrates this raw data zoom for the example source of the distant cluster RDCS J1252-29 at z =1.24. Note the point source in the upper left corner for comparison. As a means for checking artifacts introduced by one of the detectors, the contours for the individual instruments are placed on top of each other in the 2.5′ 2.5′ × image shown in panel 2. 6.3. DEVELOPMENT OF AN X-RAY REDUCTION AND ANALYSIS PIPELINE 103

The optical source classification is based on image data provided through the Second Digitized Sky Survey28 (DSS-2). Of particular interest for distant cluster applications are the red band (DSS-red) and the NIR data (DSS-NIR), which reach almost all-sky coverage and are online accessible. The imaging data is complemented by a near-position query around the source location to the NASA Extragalactic Data Base29 (NED), which contains information on all published extragalactic objects such as source type, spectral properties, and redshift, if available. The returned object list is filtered and transformed into region files, which can be overlaid onto the image data in addition to the X-ray contours. Panels 3 & 4 of Fig. 6.12 display the resulting 2.5′ 2.5′ image zooms for the DSS-red data (left) and the DSS-NIR (right). Additional 5′ 5′ overlay× images of the source environment × allow an inspection on a larger scale. Green circles indicate NED objects without redshift information, whereas red diamonds display the location of objects with the NED specified redshift. In the example of Fig. 6.12, the indicated central redshift is associated with the distant ROSAT-discovered target cluster of the observation.

6.3.5 Distant cluster optimizations The largely automated XDCP X-ray pipeline processing, as outlined in Fig. 6.8, is now completed. The two newly tested non-standard detection schemes (single band and spectral matched filter), for a complementary use to the standard source detection approach, can be considered as key characteristics. Numerous additional features of the XDCP pipeline aim at an optimization of the distant cluster detection and source classification efficiency. The efficacy of the newly implemented XDCP pipeline characteristics has been tested in initial performance comparisons to ‘standard method’ results and cross-checked with early observational feedback from follow-up imaging campaigns. In general, four categories can be distinguished for which improvement effects compared to a standard detection procedure are expected: (i) the survey usability of additional XMM fields (F), (ii) an improved extended source sensitivity (S), (iii) a higher extent reliability (R), and (iv) an optimized low-redshift screening efficiency (Z). The main high-z optimization features are summarized in the following, effect categories (F, S, R, Z) are indicated in brackets after the items:

a strict double ﬂare cleaning procedure (F+S); • correction for out-of-time events (R); • use of detection masks with excised bright sources (F+S+R); • use of an optimized distant cluster band for the detection and visual inspection (S); • use of a smooth background model ﬁt instead of spline interpolations (R+S); • four box detection iterations with increasing cell size for an improved extended source • sensitivity (S);

28The ESO DSS archive can be accessed at http://archive.eso.org/dss. 29The NED homepage is found under http://nedwww.ipac.caltech.edu. 104 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

a maximum extent radius of 20 pixels (80′′) to constrain the possible extended source • parameter space and prevent very extended spurious sources (R); maximum likelihood tests for a possible double point source blend (R); • use of detection parameter information of three diﬀerent schemes for cross-comparison • (S+R); use of the wavelet detection method for complementary extent information (R); • individual detector artefact assessment using separate X-ray source contours (R); • additional use of DSS-NIR data for an object cross-correlation with DSS-red images • (Z); automated NED object projections for source identiﬁcations (Z). • 6.4 Source Screening and Candidate Selection

The XDCP X-ray pipeline has been applied to the 546 XMM-Newton archive fields that have passed the survey definition selection (see Sect. 6.2) and a first visual screening30. These fields are listed in Tab. A.1 in the appendix with their positions, nominal and clean exposure times, galactic hydrogen column density, and the survey status. The total nominal exposure time of this data set of 17.5 Msec corresponds to 6.7 months of continuous observations, and the full X-ray reduction required about two months of CPU time. 77 of the data sets have been discarded from survey use for three possible reasons. (i) The observation was completely contaminated by soft proton flares, (ii) the central target covered most of the field-of-view, or (iii) data corruption preventing a proper processing. The remaining 469 XMM fields with a total nominal exposure time of 15.2 Msec define the XDCP core survey sample, i.e. the data sets that are in principle usable for serendipitous distant cluster searches. More than 2 000 extended X-ray sources have been detected in these survey fields. In this section, we will discuss the screening and classification procedure for these sources with the final goal of compiling a well-selected XDCP distant cluster candidate list as a starting point for further follow-up observations. The challenges associated with the detection of extended sources at faint flux levels, in particular the various reasons for the expected significant fraction of spurious non-cluster sources, have been discussed in Sect. 6.3.3. The final classification of the detected sources is too complex (see e.g. Fig. 6.10) to be reliably implemented in an automated and objective fashion. The screening process of the individual extended X-ray sources hence requires human intervention for a due assessment of possible problems and for achieving the best results. A certain level of subjectivity is inevitable at this point. However, after an initial ‘training phase’ with observational feedback (see e.g. Sect. 6.5), a relatively homogeneous classification of the sources is possible.

30Some (contaminated) XMM ﬁelds can be discarded from the survey sample based on the inspection of available quicklook data images. 6.4. SOURCE SCREENING AND CANDIDATE SELECTION 105

XDCP Candidate Selection

Selection of Extended X-ray Sources

SAS Box Detection & ML Fitting SAS Wavelet Detection

Setting 1 EXT ML Single 0.35–2.4 keV ≥2.5 σ

Setting 2 EXT ML SMF 0.3–7.5 keV ≥2 σ

Setting 3 EXT ML STD 0.3–4.5 keV ≥1.5 σ

75% +20% +5% Screen Level 0 2 000 100% Raw Extended Source List -15% Bad X-ray X-ray Catalog Cleaning Sources • FoV edge • source-in-source Screen Level 1 • high background 1 700 85% • Cleaned Extended Source List etc

-20% X-ray Source Screening Spurious Sources • spurious detections • Screen Level 2 spurious extent 1 300 65% (point-like) Screened Extended Source List

X-ray-DSS-Overlay Screening

-5% Non-Cluster DSS Optical Classiﬁcation Sources • galaxies • SNR Screen Level 3 • etc 1 200 60% Cluster Candidate Source List

Low-Interm. 15% XDCP DSS 45% Distant Cluster Identiﬁed Redshift 250 Cluster List +doubles 750 Candidates Candidates +doubles secondary science

SPT Follow-Up Observations ALL Sample XDCP Imaging Classiﬁcation Cluster Candidate 2.5% 7.5% 37.5% 7.5% Masterlist z≥1 0.5≤z<1FP5% z<0.8 FP

Figure 6.13: Source screening ﬂow chart. Used abbreviations are EXT ML: Extent Maximum Likelihood, SMF: Spectral Matched Filter, STD: Standard Method, DSS: Digitized Sky Survey, SPT: South Pole Telescope, FP: False Positives. Stated percentages refer to the Level 0 starting sample, absolute numbers are circled in blue. 106 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

6.4.1 Screening steps The complete XDCP screening and classification procedure is illustrated in the flow chart of Fig. 6.13. The selection process can be structured into the two independent parts: (i) the X-ray quality assessment of the source and (ii) the optical counterpart identification. The third part in Fig. 6.13 displays some rounded preliminary success rates of follow-up observations, covered in Chap. 7, in order to complete the start-to-end source accounting. Numbers circled in blue refer to approximate absolute extended source counts at the given screening stage, the green circled percentages are relative to the starting sample at screen level 0. The sample completeness is a prime factor for the XDCP science applications. Conse- quently care has to be taken not to lose any real clusters during the screening procedure. Ambiguous sources at a given screening stage are therefore passed to the next level for a re-evaluation before they are rejected. In the end, the survey completeness and the follow- up efficiency have to be balanced, i.e. the false positive fraction should be acceptable for a given completeness level. The source flagging and classification scheme should thus be flexible enough to allow re-adjustments of the selection cuts as a consequence of observational follow-up feedback.

X-ray selection The starting point for the selection process is the raw extended source list or screen level 0, which is compiled in a automated and objective way. The first aim of this starting source sample is the complete coverage of the single band scheme detections on the 2.5 σ extent significance level, which contribute about 75% (see Fig. 6.13). The second aim is to test for possible selection effects introduced by a given detection scheme. We thus include the extended sources detected with the spectral matched filter scheme on the 2.0 σ significance level which are not on the list and comprise about 20% of the initial sample. Below an extent significance of 2.0 σ most additional detected sources are very uncertain. For the standard scheme results at the 1.5 σ extent level, only those new sources are added to the list, which are also confirmed as extended by the wavelet method. With these last 5% of sources detected at lower significance, the raw extended source list with a total of just over 2 000 extended sources is completed. From this point on, all subsequent classification steps (screen levels 1–3 ) rely on human intervention. As a first screening level, the raw X-ray source catalog is cleaned of obvious spurious detections based on the summary plots as shown in the upper right panel of Fig. 6.10. At this time, the majority of the source-in-source detections, FoV boundary sources, wing detections of bright point sources, and pronounced out-of-time residual detections are removed. With about 15% of the X-ray sources flagged as ‘artifact detections’, the cleaned screen level 1 sources are passed on two the next level. The detailed X-ray source screening of the second level makes use of the zoomed X-ray data overlays as shown in panels 1 & 2 of Fig. 6.12 and (up to) three sets of returned source parameters of the maximum likelihood fitting procedure. At this stage, sources with more 6.4. SOURCE SCREENING AND CANDIDATE SELECTION 107

subtle reasons for a spurious extent or a spurious detection are identified. The most common causes are misclassified point sources, point source blends, high noise environments, poorly determined backgrounds, and detections at high off-axis angle due to PSF residuals. Besides the visual confirmation of the existence and extent using the overlay panels 1 & 2, the source quality is evaluated based on (i) the wavelet detection result, (ii) the number of extent detections and their significance within the three schemes, (iii) the stability of the extent determination, and (iv) the hardness ratio of the source. Spurious extents of point sources can often be identified based on cross-checks with items (i), (ii), and (iv); spurious detections on the other hand typically do not exhibit stable extent solutions of item (iii) or are not re-detected at all for tests (i) and (ii). An additional 20% of the initial source sample are identified as spurious sources in this way and flagged∼ accordingly. Extended sources that passed the X-ray quality screening but are close to the cut threshold are marked as sources with lower X-ray significance and retained in the object list. The screen level 2 source list has been cleaned of about 35% of the initial sample. This large fraction of spurious sources is to be predominantly attributed to systematic effects inherent to the available SAS source detection and calibration tools and not to the expectation rate of false positives of the statistical likelihood. The main reason for the built-in redundancies in the XDCP source detection procedure is the improved identification and removal of this spurious source fraction while retaining the full sensitivity. The screen level 2 extended source list completes the X-ray object selection and characterization and should now closely reflect the statistical expectations of a sample with a 2–2.5 σ significance threshold.

Optical classification The screened extended source list is the input for the optical classification, which constitutes the second part of the candidate selection process. The main tools for the optical counterpart search are the X-ray-DSS overlays for each source as shown in panels 3 & 4 of Fig. 6.12. As a first step of the optical classification, the non-cluster sources (e.g. single nearby galaxies) are visually identified using the DSS imaging data and removed from the sample. The remaining 60% percent of the input sample comprise the screen level 3 cluster candidate source list of roughly 1 200 objects, still including multiple detections of the same source in different fields. In the last step, these sources are visually classified either as DSS-identified cluster candidates or as XDCP distant cluster candidates based on their optical counterpart or the lack thereof. The classification scheme will be discussed in the following Sect. 6.4.2. About three quarters of the screen level 3 cluster candidates (45% of the initial sample) can be identified with an optical cluster counterpart signature on the X-ray-DSS overlays. These approximately 750 distinct objects are flagged as low–intermediate redshift cluster candidates and are saved for secondary science projects beyond the XDCP scope. The 250 extended X-ray sources without an DSS-identified optical counterpart enter ∼ the distant cluster candidate master list, which constitutes the basis of the subsequent follow-up imaging program. In order to provide a full accounting of the extended X-ray source population at the XDCP sensitivity level, approximate cluster yields are stated in 108 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Fig. 6.13 as derived from the follow-up imaging programs of Chap. 7. The numbers related to the DSS-identified candidates are based on a preliminary evaluation of the wide-field follow-up program in the South Pole Telescope (SPT) survey region (see Sect. 9.2 & 12.1). The first important finding from these data is the confirmation that the sources flagged as spurious are indeed not associated with optical clusters out the data limit of z 0.9. ∼ The final cluster accounting establishes the following approximate cluster yields for the distant cluster candidates: About one out of six candidates is a true z > 1 system, half of the sources are associated with z< 1 clusters, and roughly a third are∼ false positives (FP), i.e. spurious sources. For the DSS-identified candidate sources, about 5/6 are z <0.8 clusters or groups with a false positive rate of 1/6. ∼ The XDCP candidate selection task (including spurious sources) can hence be globally summarized as follows (see Fig. 6.13). (i) About 1/40 of all extended X-ray sources are associated with the targeted population of z >1 galaxy clusters. (ii) 1/10 of all extended ∼ sources are clusters beyond the DSS identification∼ limit. (iii) 1/2 of all extended sources are galaxy clusters. (iv) 3/4 of all cluster sources can be identified∼ with an optical DSS ∼ counterpart. Concerning the required cleaning of spurious sources, the following picture is obtained. (v) For each distant cluster candidate more than two spurious sources were removed from the initial sample in early cleaning stages. (vi) 3/4 of all spurious sources can be identified as such within the screening procedure without∼ follow-up observations. (vii) 1/10 of all spurious sources pass the screening procedure and need to be identified as false∼ positive sources in the follow-up data of the distant cluster candidate sample. (viii) After removing all spurious sources from the object catalog, > 90% of the non-spurious extended sources in the high galactic latitude survey fields are∼ associated with galaxy clusters. Additional source diagnostic plots will be presented in Sect. 6.6.

6.4.2 Classification scheme After the general overview of the screening procedure, we will now have a closer look at the cluster candidate classification scheme. As discussed, the selection process is comprised of the two independent parts of the X-ray quality assessment of the source, followed by the subsequent evaluation of possible optical counterparts on DSS imaging data. These two screening dimensions are reflected in the classification scheme illustrated in the left panel of Fig. 6.14. In a schematic way, the optical DSS visibility is plotted against the X-ray quality of the source. Beyond the X-ray significance and quality threshold (vertical dashed line), sources are flagged as spurious (red region). The DSS limit (horizontal dashed line) divides the plane into the distant cluster candidates in the upper half (green) and DSS-identified systems in the lower half (blue). In this scheme the ideal distant candidates, featuring excellent X-ray quality in combination with a blank DSS sky region, are located in the upper left corner. Approaching the DSS limit, i.e. moving downwards, results in traces of possible cluster galaxies at a level insufficient for a secure cluster identification and towards the right side of the plot, the X-ray significance of the extended sources is reduced. 6.4. SOURCE SCREENING AND CANDIDATE SELECTION 109

best best candidates cand. cand. good with lower cand. X-ray quality DSS good spurious clusters candidates & point candidates with object traces sources candidates DSS DSS with clusters DSS groups object candidates & groups

clusters traces with lower cluster mass DSS & groups X-ray clusters DSS quality

decreasing DSS visibility groups

decreasing X-ray quality redshift Figure 6.14: Qualitative X-ray source flagging scheme. Left: Practical implementation in the DSS-object-visibility versus X-ray-quality plane. The dashed lines indicate qualitatively the DSS limit (horizontal) and the X-ray quality limit (vertical) beyond which sources are classified as spurious or point-like. Right: Qualitative correspondence of the flagging scheme in the cluster mass versus redshift plane for ideal clusters. The dashed line represents the DSS identification limit.

Figure 6.15: Candidate classification prototypes (2.5′ 2.5′ FoV) following the scheme in the left × panel of Fig. 6.14. Top row: Distant cluster candidates of best quality (left), with slight object traces (center) and with weaker X-ray significance (right). Bottom row: DSS-identified cluster and group sources with high confidence (left), medium-high confidence (center), and a DSS-limit case (right) requiring the cross-comparison of the optical and NIR DSS data for enhancing the object likelihood of weak traces. The NED label in the upper left panel identifies XMMU J2235.3-2557 at z =1.39. 110 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Relating DSS visibility qualitatively to redshift and X-ray quality to cluster mass, the classification scheme is transformed, in first order, to depend on the prime cluster characteristics (right panel of Fig. 6.14). The diagonal dashed line indicates the DSS limit which is pushed towards higher redshift for increased cluster masses. Increasing the redshift of a system or decreasing the mass results on both cases in a lower X-ray ‘quality’ of the cluster candidate (i.e. a lower signal). In general, the qualitative classification dimensions DSS visibility and X-ray quality are related to the physical cluster characteristics redshift and mass in a more complex way. Besides the mass and redshift dependence, the observed X-ray significance and quality of the candidate is influenced by (i) the effective exposure time at the source location, (ii) the source extent, (iii) the level of the X-ray background, (iv) the galactic hydrogen column at the source location, (v) nearby contaminating sources in the field, (vi) AGN emission in the cluster, or (vii) ongoing merging activity in the system. The DSS counterpart identification, on the other hand, is additionally influenced by (i) the magnitude and location of the brightest cluster galaxies, (ii) the local DSS depth, (iii) an increasing foreground object density at lower galactic latitude, (iv) contamination from optically bright nearby objects, and (v) the lack of DSS-NIR data in some sky regions. Figure 6.15 illustrates the classification prototypes following the discussed scheme. The classification is applied to all cluster candidates of the screen level 3 list. The prototypes in the upper row show distant cluster candidates, whereas in the bottom part examples for DSS-identified systems are displayed. In the next section, we will address the question of the DSS identification limit in some more detail.

6.4.3 DSS identification limit The image data of the second generation Digitized Sky Survey (Reid et al., 1991) is based on photographic plate scans of several wide-field surveys using 1 m-class Schmidt telescopes. The typical exposure times for the DSS-red plates, corresponding roughly to the R filter band (see Tab. 7.1), are 60-70 min. The DSS-NIR plates have been exposed for 80-90 min and with a central wavelength of λc 8500 A˚ this band is similar to the (optical) I band. The average limiting point source sensitivities∼ (Vega magnitudes) are R 20.8 mag (DSS- lim ≃ red), and Ilim 19.5 mag (DSS-NIR) with local variations of about 0.4 mag for sky regions with the deepest≃ and shallowest coverage. The limit in the R-band± for the DSS-red data corresponds to the expected apparent magnitude of an M* elliptical galaxy at redshift z 0.5, an M* 1 elliptical at z 0.65, or an M* 2 object at z 0.8 (see Fig. 7.1 and Tab.∼ 7.2). For the− I-band of the DSS-NIR∼ images the− corresponding∼ redshifts are z 0.45 ∼ for an M* galaxy, z 0.65 for M* 1, and z 0.85 for an M* 2 object. The brightest cluster∼ galaxies− (BCGs) are∼ typically 1–2 mag− nitude brighter than the characteristic luminosity L* (see Sect. 11.1). This implies that BCG signatures should be generally visible in DSS-NIR images out to redshifts of z 0.5–0.8. The significance of DSS sources close to the limiting magnitude can be additionally∼ enhanced by cross-comparing the DSS-red and DSS-NIR images for spatially coincident object traces. 6.4. SOURCE SCREENING AND CANDIDATE SELECTION 111

Figure 6.16: DSS identiﬁcation limit for rich clusters. Marked object positions are either associated with color selected cluster members (circles) with shown H-band Vega magnitudes or spectroscopically conﬁrmed cluster members (squares and diamonds). All images are 2.5′ 2.5′ in size. × 112 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Figure 6.16 shows a compilation of six spectroscopically confirmed galaxy clusters with redshifts 0.2 < z < 0.8. The corresponding 2.5′ 2.5′ DSS-red (left) and DSS-NIR images × (center) are displayed∼ ∼ with high-contrast color cuts around the noise limit of the data. The right panels present NIR color composites taken, if available, from the follow-up imaging data of Chap. 7 to indicate the underlying deep optical cluster signature. The overlaid blue circles on the DSS data represent color selected cluster members with associated H-band magnitudes. For the two systems without additional follow-up data, the sky positions of spectroscopically confirmed cluster members are marked. As can be seen from Fig. 6.16, the optical DSS-identification of clusters out to z 0.4 (green box) is straightforward and is generally possible at high confidence levels. In∼ the redshift range 0.40.6 the DSS-based cluster identification becomes increasingly difficult and is beyond the DSS capabilities in the general case. However, for favorable combinations of BCG properties, cluster richness, and local DSS depth, the optical counterpart can, in some cases, be identified out to z 0.8. Cl1216 12 exhibits weak traces of about half a dozen probable cluster galaxies, whereas∼ for Cl0152− 13 only one probable AGN counterpart is − securely detected and some weak indications of the BCG are visible in the DSS-NIR image. We conclude that the average DSS cluster identification limit is in the redshift range z 0.5–0.6, if the full depth of the DSS-red and DSS-NIR images is exploited. This limit is≃ also confirmed by early XDCP selection performance tests in the COSMOS field as presented in the next section.

6.5 Performance Tests in the COSMOS Field

The XDCP X-ray pipeline of Sect. 6.3 and the candidate selection procedure of Sect. 6.4 has been tested in its early version in the COSMOS survey field (see Sect. 4.4). The COSMOS data set31 is particularly well suited for a performance benchmark since (i) the individual XMM pointings are of similar depth as the average XDCP survey fields, and (ii) the multi-wavelength coverage of the field enables a direct cross-comparison between detected extended X-ray sources and galaxy overdensities with corresponding accurate photometric redshifts. The XDCP reduction and selection machinery has been (blindly) applied to 20 individual XMM fields of the first COSMOS X-ray data set, corresponding to about 85% of the observations required to cover the full 2.1 deg2 field. The average cleaned exposure time of 24.9 ksec per field is about 30% deeper than the median coverage of the XDCP fields. The fields are overlapping to contiguously map the 1.4◦ 1.4◦area. In contrast to the × COSMOS X-ray analysis of Finoguenov et al. (2007), which made use of the reconstructed mosaic image with the full co-added effective exposure at each position, the XDCP test

31Only the ﬁrst XMM data set has been used, meanwhile the available COSMOS X-ray coverage has more than doubled in exposure time. 6.5. PERFORMANCE TESTS IN THE COSMOS FIELD 113

XMM Distant Cluster Search 600

500

187 112 41 400 (202) (102) (19)

300 dn/dz

200

100

0 0.0 0.5 1.0 1.5 2.0 z

Figure 6.17: Left: Redshift distribution of identified clusters and candidates in the COSMOS field (black line). The blue dotted line shows the distribution of DSS-identified clusters in redshift bins of ∆z = 0.2, the green dashed line represents the distant cluster candidates. The photometric redshift estimates are based on early internal versions of the COSMOS galaxy catalog. The bump at z 1 is ∼ also present in the published cluster catalog (Finoguenov et al., 2007) and is still under investigation. The five candidates at z< 0.6 were identified as low mass groups, which are harder to identify on DSS images. Right: The expected shape of the redshift distribution for an XDCP-like survey without cluster evolution (solid line) and with evolution (dashed line). Plot from C. Mullis. procedure was applied to the individual XMM fields as for the real survey fields. Following the compilation of the final screen level 3 source sample with distant cluster candidates and DSS-identified systems, the XDCP object list was cross-correlated with the proper COSMOS X-ray cluster list, which is based on the deeper mosaic data in conjunction with the photometric confirmation based on the optical data. The left panel of Fig. 6.17 illustrates the results of this test in form of the binned photometric redshift distribution of all 46 XDCP identified clusters (black line). The blue dashed line represents the DSS-identified clusters, whereas the green line indicates the selected distant cluster candidates. Six distant z > 1 clusters with X-ray fluxes down to 4 10−15 erg s−1cm−2 [0.5–2.0 keV] were successfully∼ identified. The five candidates at the lower∼ × redshift end at z < 0.6 (green line) have been revealed as lower mass groups, mostly ∼ 13 in the range M 3–4 10 M⊙. Due to their lower optical richness, these systems are harder to distinguish≃ from× possible foreground objects. The blue line for the DSS-identified clusters, on the other hand, confirms the average DSS redshift limit of z 0.5–0.6. The ≃ fraction of false positive sources in the XDCP sample turned out to be 6/22 (27%) for the distant cluster candidates, and 2/32 (6%) for the DSS flagged objects. A slightly worse performance for the real XDCP survey data can be expected due to more contaminations in serendipitous fields, which have not been specifically selected for deep extragalactic survey applications. The right panel of Fig. 6.17 shows the expected shape of the cluster redshift distribution for an XDCP-like survey, derived for early sensitivity estimates. The solid line represents a model without cluster evolution, whereas the dashed line includes evolution effects as expected from the X-ray luminosity function analysis of Fig. 4.3. 114 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Finoguenov et al. (2007) find no substantial evolution in their analysis of the initial 72 cluster COSMOS sample out to z 1.3, selected from the first 36 XMM pointings. The ∼ COSMOS cluster luminosities cover the range 8 1042–2 1044 erg s−1 and hence do not include the high mass end for which mild evolution× effects hav× e been established (Sect. 4.3). The pronounced bump in the observed redshift distribution around z 1 (left panel), also ∼ present in the published COSMOS data and still under investigation, is likely attributed to either physical large-scale structure or systematic effects in the derived photometric galaxy redshifts. In summary, the XDCP X-ray pipeline performance test in the COSMOS field has confirmed (i) the high achievable detection sensitivity, (ii) the good identification efficiency of low–intermediate redshift DSS clusters, and (iii) a high fraction of probable z > 1 systems in the distant cluster candidate sample, giving additional support to the∼ selection procedure.

6.6 Field and Source Diagnostics

In the last section of this chapter on XMM archival X-ray data, we will have a closer look at some selected statistical properties of the survey ﬁelds and the detected extended sources.

6.6.1 XMM data yield and field statistics The 540 reduced fields of the first 4.5 years of XMM operations give the possibility to trace the XMM science data yield over the course of a year and to assess the data losses due to flare periods (see Sect. 6.3.2). Figure 6.18 illustrates the seasonal changes of the science- usable data fraction. The upper panel shows the instrument-weighted clean imaging time fraction with respect to the nominal exposure time for all fields that passed the flare removal procedure. In effect, this represents the total archival imaging data percentage, and has an average value32 of 57%. The lower panel displays the absolute clean and lost times for the fields with all instruments in imaging mode. The resulting mean clean time fraction of 67% is representative for the overall XMM science data yield. Hence one third of the nominal pointing time is on average lost, predominantly due to soft proton flares and a smaller fraction due to instrument overheads. Both representations show an increased lost data fraction during the summer months and a higher data yield during winter with differences of about 20% of the total time. These pronounced seasonal space weather effects are thought33 to be related to the solar cycle and the XMM orbit evolution caused by external perturbations. The latter effect can also introduce long term changes from year to year.

32The 10% diﬀerence compared to the overall XMM science data yield is due to the fact that the full sample of 540 ﬁelds contains observations, in which some instruments were not operated in imaging mode, i.e. science modes not usable for the XDCP survey. 33See longterm compilations at http://xmm.vilspa.esa.es/docs/documents/USR-TN-0014-1-0.pdf. 6.6. FIELD AND SOURCE DIAGNOSTICS 115

Figure 6.18: Seasonal changes of the clean time fraction. Top: Average clean time fraction versus the time of the calender year in bins of two weeks. The data points are based on 540 processed XMM archive fields observed within a 4.5 year period between January 2000 and August 2004. The dashed horizontal line represents the sample average of the weighted effective clean imaging time fraction of 57%. In contrast to most optical telescope sites, the space weather is favorable in terms of data yield in the winter months between October and February. Bottom: Absolute clean exposure times (blue) and flared periods (red) in two week bins for the 439 fields for which all three detectors have been operating in imaging mode. The average clean time fraction is now 67% and is representative for the overall XMM science usable data fraction. 116 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Figure 6.19: XDCP survey field statistics. Top: Cumulative distribution of nominal exposure times (blue line) and flare cleaned times (red line) for the 469 usable XDCP survey fields. The average clean exposure time of 18.78 ksec is indicated by the vertical dotted line, the median of 15.71 ksec is illustrated by the dashed line. Bottom: Cumulative distribution of the galactic hydrogen column for the survey fields. 6.6. FIELD AND SOURCE DIAGNOSTICS 117

The statistical properties of the 469 fields of the XDCP core survey sample are summarized in Fig. 6.19. The top panel illustrates the cumulative distribution of the resulting clean exposure times (red) in comparison to the nominal distribution (blue). The average clean XDCP field depth is 18.78 ksec (dotted vertical line), the median is 15.71 ksec (dashed line). The lower panel shows the cumulative field distribution as a function of the galactic hydrogen column N with a median value of 3.55 1020 cm−2. H × 6.6.2 Extended source diagnostics We now turn to the properties of the extended X-ray sources and their detection efficiency. Figure 6.20 provides a global summary of all extended X-ray sources detected with the single band detection scheme. Plotted are the maximum likelihoods for the extent of the source (EXT ML) versus the total source counts in the 0.35-2.4 keV band. The horizontal cut-off at EXT ML=5 indicates the lower significance threshold for the detection run. Distant cluster candidates are represented by dark blue circles. The median number of total source counts for this population is 193, the median EXT ML is 11.7, corresponding to 5–6 σ significance, and the median detection likelihood amounts to 36.6. The lighter blue circles indicate the population of DSS-identified cluster candidates. Red symbols illustrate the distribution of sources classified as spurious extent (plus signs) or spurious detections (crosses). The extent maximum likelihood exhibits a decent correlation with the total photon number of the cluster candidates (blue symbols), with a scatter of about a factor of two. The extended sources classified as spurious show on average a lower EXT ML for a given total number of source counts. Even though spurious sources show a systematic offset in their EXT ML, they still occupy almost the full parameter space populated by cluster sources, in particular at the important fainter levels. The first important conclusion from this global view on the detected extended source population is that spurious sources cannot be automatically separated from real clusters by means of appropriate parameter cuts. This is to be attributed to the discussed systematic (calibration) effects that dominate the spurious detections for the eboxdetect and emldetect methods. The second preliminary conclusion is that the sensitivity goal for extended sources down to a total number of 100 counts and less seems to be achievable. The typical significance at which objects with 100 X-ray photons can be resolved into ∼ extended sources is about 4 σ. The final survey sensitivity limit has to be quantified based on detailed simulations (see Sect. 9.1.3). As a next step, we can compare the extent likelihoods of the sources for the three different detection schemes, which is shown in Fig. 6.21 for the distant cluster candidate sample. The EXT MLs of the standard scheme are plotted along the abscissa, whereas the corresponding source likelihoods for the single band scheme (dark squares) and spectral matched filter scheme (light blue circles) are represented by the ordinate. The likelihoods for the single band scheme and standard scheme scatter along the lower red line which indicates matching values. The corresponding comparison between the schemes with a wider spectral coverage exhibits a slightly tighter correlation about the upper red line, which indicates the average offset factor of 2.8 and corresponds to the 118 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

Figure 6.20: Extent likelihood versus X-ray source counts. Dark blue circles represent distant cluster candidates, light blue circles DSS-identified clusters, and red symbols sources that have been classified as spurious detections. The lower horizontal line indicates the minimum required extent likelihood, whereas the diagonal line shows the general correlation for the good sources. Sources that have been classified as spurious tend to have a lower extent likelihood compared to clusters candidates with the same number of counts. However, the populations are still largely overlapping which prevents an automated identification of false detection. mean effective weight. The effective weighting factor of the spectral matched filter scheme is not fixed a priori, but depends on the spectral properties of the source. Below standard scheme EXT MLs of about 8, the distant cluster candidates appear to be systematically above the average correlation which points towards an achieved enhancement of the likelihoods at faint flux levels. The global statistics for the spectral matched filter detections confirm that the average likelihood offsets with respect to the standard scheme are increased by about 6% for distant cluster candidates compared to sources classified as spurious. However, no firm conclusion on the effectiveness of the spectral matched filter scheme can be drawn before a significant number of distant candidate clusters at lower EXT ML levels have been spectroscopically confirmed. Last but not least important, the optimal detector area for serendipitous surveys is investigated. As discussed in Sect. 6.1, XMM-Newton’s grasp as survey instrument increases with the maximum off-axis angle Θ. On the other hand, the vignetting effect (Fig. 6.2) decreases the sensitivity towards the outskirts, and more importantly, the PSF shape (Fig. 6.3) broadens and becomes increasingly ill-behaved at large Θ, which hampers the detection of extended sources. 6.6. FIELD AND SOURCE DIAGNOSTICS 119

Figure 6.21: Extent likelihood comparison of the three detection schemes. The dark squares display the extent likelihoods of the single 0.35-2.4 keV detection band versus the EXTMLs of the standard detection scheme for distant cluster candidates with the lower red line indicating matching likelihoods. The top line and the lighter blue circles show the situation for the spectral matched filter scheme which exhibits an offset of roughly a factor of 2.8 corresponding to the average effective weight.

Figure 6.22: Cluster detections per unit sky area as a function of off-axis angle normalized to the detections at 10′ off-axis angle (dashed red line). The detection rate is almost constant between 7′ and 12′ from the optical axis. In the central region the detection rate rises owing to the increased effective area. Beyond 12′ the increasing vignetting and the broadened PSF decrease the cluster detections per unit area. The inner 12′ (solid red line) are well-behaved and can be used for the core survey. 120 CHAPTER 6. X-RAY ANALYSIS OF XMM-NEWTON ARCHIVAL DATA

One way to address the issue of the maximum off-axis angle for a strictly selected survey sample is to ask at what point the cluster detections per unit sky area start dropping significantly. Figure 6.22 shows the results obtained from 120 fields without masked regions, i.e. no large foreground objects, and a total of 375 cluster sources (screen level 3 objects). The plot traces the normalized cluster detections per unit sky area as a function of off- axis angle, which is derived from the cumulative distribution N(< Θ) divided by the enclosed sky area Ω(< Θ) and normalized to the average value within 10′ radius (dashed red line). From Fig. 6.22 it can be seen that the cluster detections per unit area are almost constant between off-axis angles of 7′ <Θ<12′. At larger radii the detection rate decreases ∼ ∼ ′ significantly. Hence, a maximum off-axis angle of Θmax =12 promises to be the best compromise between survey grasp and completeness. With this cut-off radius for the final 2 2 XDCP core sample, the XMM survey grasp is g12 204 cm deg , a 30% improvement with respect to a 10′ constraint. ≃

6.7 Summary

At the end of the technical chapters 6–8, the main contributions and achievements of this thesis work towards the overall XDCP survey project are brieﬂy listed. Concerning the X-ray analysis of XMM-Newton archival data this work can be summarized as follows:

Deﬁnition of the current XDCP survey ﬁeld sample based on a screening of the full • XMM archive;

Development of a distant-cluster-optimized X-ray pipeline for the data reduction, • source detection, and creation of X-ray optical overlays;

Conduction of performance tests for different detection schemes; • Tests of the full XDCP selection procedure in the COSMOS field; • Reduction of 546 XMM observations with a complete analysis of 469 XDCP suitable • survey fields;

Screening and classification of about 2 000 detected extended sources; • Compilation of a final XDCP master source list with 250 distant cluster candidates • and about 750 additional DSS-identified cluster candidates∼ at low or intermediate redshifts. Chapter 7

Near-Infrared Follow-up Imaging

This chapter will deal with the follow-up imaging of the X-ray selected, distant cluster candidates, i.e. with steps 3 and 4 of the survey strategy discussed in Sect. 5.2. Although the R–Z method used for the pilot study resulted in the discovery of XMMUJ2235.3-2557, certain restrictions and limitations gave rise to the motivation to develop an alternative follow-up strategy. The following sections will discuss in detail the theoretical expectations of a NIR-based follow-up approach, its practical implementation, and the required software and analysis tools.

7.1 A New NIR Imaging Strategy

7.1.1 Motivation A new two-band imaging approach should ideally cure both the practical restrictions related to the observatory site and telescope access and the intrinsic systematic limitations of the R–Z method. Wanted is thus an imaging strategy that: 1. is applicable to 4 m-class telescopes and is thus not dependent on the VLT with its high over-subscription rate; 2. is extendable to the Northern sky for a future enlargement of the XDCP survey area; 3. yields more accurate redshift estimates in the main target range 1.0

121 122 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

Figure 7.1: Simple stellar population (SSP) models for passively evolving galaxies.Top panels: Apparent magnitudes of L* galaxies (Vega system) versus redshift diagrams for a formation redshift of zf =5 for the optical and NIR bands V, R, I, Z, J, H, and Ks. Center left: Color evolution diagram for a selection of colors (zf =5). Center right: R–Z, Z–H, and I–H color shifted to the same origin at z=0.5. The slope of the relations determine the redshift sensitivity in the different redshift regimes. Bottom panels: Model grid of the R–Z (left) and Z–H (right) color evolution for formation redshifts of three, five, and ten (different colors) and solar (solid lines) and three times solar metallicity (dashed lines). Models computed by Daniele Pierini using PEGASE2 (Fioc and Rocca-Volmerange, 1997). 7.1. A NEW NIR IMAGING STRATEGY 123

Filter Center Cut-on 5% Cut-oﬀ 5% mAB (Vega) D4000 D4000 [µm] [µm] (zstart ) [µm] (zend ) 1 Rspecial 0.655 0.53 (0.3) 0.75 (0.9) 0.195 I2 0.798 0.70 (0.8) 0.89 (1.2) 0.440 Z3 0.90 0.81 (1.0) 0.99 (1.5) 0.521 J 1.22 1.09 (1.7) 1.36 (2.4) 0.857 H 1.64 1.49 (2.7) 1.81 (3.5) 1.372 Ks 2.15 1.97 (3.9) 2.33 (4.8) 1.855

Table 7.1: XDCP filters of interest. For the three optical filters R, I, and Z and the standard NIR bands J, H, and Ks the central wavelength and the 5% transmission start and end points are listed. The values in brackets indicate the redshift when the D4000 break enters (zstart) and exits (zend) the filter band. The last column shows the offsets between the standard photometric AB-system widely used in optical bands, and the classic Vega system which still prevails in NIR observations. The relation mAB =mVega+mAB(Vega) transforms the magnitudes.

7.1.2 Imaging strategy The natural extension for distant cluster imaging beyond the reddest suitable optical combination of filters, R–Z, is to completely move into the near-infrared regime. Table 7.1 summarizes the characteristics of the reddest optical (standard) broad band filters (R, I, Z4) and the three standard NIR bands J, H, and Ks5. Near-infrared imaging has only recently become a competitive alternative to optical observations and has been strongly boosted by great advances in the NIR imaging array technology. The latest generation of 2k 2k NIR arrays have been available for about 5 years now and have driven the development× of a new generation of sensitive wide-field NIR instruments. For a first approach to the problem, we can assume that suitable instrumentation exists for all six filter bands of interest and we can thus restrict ourselves to physical arguments related to the targets and the restrictions imposed by ground-based observations in general. A discussion of the practical implementation will follow in Sect. 7.2. Secondly, the discussion will focus on the properties of the ubiquitous elliptical galaxy population in clusters, as introduced in Sect. 2.4. As ‘red and dead’ objects, the spectral energy distribution (SED) of elliptical galaxies can nowadays be accurately modelled and predicted with simple stellar population (SSP models) codes such as PEGASE2 (Fioc and Rocca-Volmerange, 1997) or GISSEL (Bruzual and Charlot, 2003). The SSP models under consideration assume an instantaneous star formation burst at a given formation redshift

1For FORS2 R band see http://www.eso.org/instruments/fors/inst/Filters. 2For EMMI I band see http://www.ls.eso.org/lasilla/sciops/ntt/emmi/emmiFilters.html. 3For Z, J, H, Ks filters see http://www.mpia.de/IRCAM/O2000/INFO/FILTERS/Filter-Status.html. 4In order to avoid confusion with the redshift z, the SDSS broad-band filter termed z is capitalized and denoted as Z throughout this thesis. 5Ks is a modified ‘short’ K filter used by the 2MASS survey to reduce the thermal background at the long wavelength end. 124 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

Filter Sky m*(z=0.5) m*(z=1.0) m*(z=1.5) m*(z=2.0) ∆m1→2 [mag asec−2] [Vega (AB)] [Vega (AB)] [Vega (AB)] [Vega (AB)] [mag]

Rspecial 19.9–21.1 20.9 (21.1) 23.6 (23.8) 25.7 (25.9) 27.0 (27.2) 3.4 I 19.2–19.9 19.8 (20.2) 22.2 (22.6) 23.8 (24.2) 25.4 (25.8) 3.2 Z 18.1-18.8 19.4 (19.9) 21.4 (21.9) 23.3 (23.8) 24.2 (24.7) 2.8 J 15.8–16.7 18.5 (19.4) 20.2 (21.1) 21.2 (21.9) 22.2 (23.1) 2.0 H 13.8–15.0 17.5 (18.9) 19.1 (20.5) 19.9 (21.3) 20.5 (21.9) 1.4 Ks 12.7–13.0 16.9 (18.8) 18.2 (20.1) 18.9 (20.8) 19.6 (21.5) 1.4

Table 7.2: Benchmark of the apparent magnitude evolution of L* passive galaxies with formation redshift zf = 5 and solar metallicity for various ﬁlter bands. The second column states the typical sky surface brightness as measured for the ESO La Silla site7. The given intervals span bright time (ﬁrst value) and dark time (second value) in the Vega system. For faint galaxy observations with long exposure times, the assumption of background limited observing conditions is valid for all bands of interest. Columns 3–6 show the model expectations for the apparent galaxy magnitudes at various redshifts. The last column summarizes by how many magnitudes the galaxies drop between redshift 1 and 2, indicating that for the bluer bands even the largest telescopes will start losing the objects between z 1–1.5. ∼

zf and take a stellar initial mass function (IMF), the total stellar mass, and the rate of chemical enrichment as input. With this it is possible to compute the age-dependent distribution of stars in the Hertzsprung-Russell (HR) diagram and thus the properties of the stars contributing to the integrated light. Using this evolutionary population synthesis technique (Tinsley, 1978), the full SED of the passively evolving galaxies can be obtained at any time t after the burst by summing the contributions of the different stellar populations at this age using a stellar spectral flux library (Pickles, 1998). 7 The prediction of these simple stellar population models for a formation redshift zf =5 are summarized in the top panels of Fig. 7.1, where the apparent magnitudes in various bands are shown for passively evolving galaxies with characteristic absolute magnitude M . The derived color evolution for a variety of suitable colors are given in the left central ∗ panel. The illustrated (red) colors show mostly a well behaved monotonic functional form in the redshift range up to z 2, which qualifies them in principle as redshift estimators. ∼ The redshift sensitivity, however, depends primarily on the slope of the color evolution (d(X Y)/dz)−1, where (X Y) denotes a given color. This sensitivity is emphasized in the right− central panel for the− color combinations I–H, Z–H, and R–Z, which are normalized to a common color origin at redshift z =0.5, since this is the fiducial XDCP starting redshift beyond which follow-up imaging is needed. Note the flattening of the R–Z color at z 1 and with a smaller effect at z 1.2 for I–H and at z 1.5 for Z–H. This feature occurs∼ ∼ ∼ 6See http://www.eso.org/gen-fac/pubs/astclim/lasilla/l-vanzi-poster, http://www.ls. eso.org/lasilla/Telescopes/2p2T/D1p5M/misc/SkyBrightness.html, and http://www.eso.org/ observing/etc/doc/ut2/uves/helpuves.html. 7PEGASE2 models computed by Daniele Pierini. 7.1. A NEW NIR IMAGING STRATEGY 125

when the 4 000 A˚ break is completely redshifted beyond the cut-off (see Tab. 7.1) of the blue filter, in this case R, I, or Z. The small amount of flux in ellipticals below 4 000 A˚ (see spectra in Fig. 4.5) is far less redshift dependent compared to the break passing through the filter. The second important parameter influencing the slope of the color evolution is the formation redshift of the galaxy. The characteristic 4 000 A˚ break of ellipticals only emerges when the young, hot, and blue stellar populations with their dominant UV flux have disappeared from the main sequence (main sequence turn-off), i.e. the flux break will evolve with the age of the stars. A younger stellar population at the epoch when the galaxy is observed, i.e. a lower formation redshift, will thus shift the observed color towards the blue. This bluing effect increases with decreasing galaxy age at the time of observation and is due to the rapid evolution of young stellar populations. The influence of the formation redshift zf on the color evolution is depicted in the bottom panels of Fig.7.1 for the R–Z and Z–H colors. The model grid also illustrates the effect of increasing the metallicity, which makes up the final critical input parameter for the predictions. It can be seen that to first order an increasing metallicity results in a redder overall color, i.e. the color evolution curves are shifted upwards as a whole. Looking at epochs with a still relatively young stellar population, i.e. at z 1.5, the age-metallicity degeneracy becomes obvious in this illustration as increasing age∼ and metallicity have similar effects on the color. This degeneracy can be disentangled with a long redshift baseline as shown in Sect. 10.1. The pre-requisites for measuring accurate galaxy colors are sufficiently deep exposures in each of the filter bands. Table 7.2 gives a benchmark summary of the predicted m* apparent magnitudes for various redshifts as shown in Fig. 7.1. The second column of Tab. 7.2 lists typical levels of the sky surface brightness in the different bands as reference. Comparing these values in the J, H, and Ks-bands to the expected magnitudes of the targeted galaxies illustrates the main challenge of ground-based NIR observations, namely that the sources of interest are between a factor of 100 (∆m =5) and 1000 (∆m =7.5) fainter than the sky background. For the optical R, I, and Z-bands, observations of distant galaxies are still background- limited (b s) . Under these observing conditions the signal-to-noise ratio (SNR) for a ≫ 8 source with signal s(texp) and total background b(texp) enclosed in the source aperture will increase with exposure time texp as

s(t ) SNR = exp t . (7.1) ∝ exp b(texp) q q For a fixed minimal SNR requirement the flux at the detection limit thus follows f (t) lim ∝ 8 The Poisson statistics with standard deviation σb = √b is applicable to the total number of detected electrons and not directly to the digital count units. For a proper statistical treatment the signal and background have to be transformed to these physical units first by multiplying with the gain g, i.e. the electrons-per-count conversion factor. Relation 7.1 further assumes that the dark current and read noise are small compared to the background noise, which is well-justified from Tab. 7.4 for the NIR. 126 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

1/√texp. Expressing this relationship in magnitudes yields ∆mlim =1 tdeep =6.3 tshallow, i.e. going just one magnitude deeper requires more than six times the→ initial exposure· time. For a doubling of the exposure time, the gain at the magnitude limit is ∆mlim =0.38 mag. These simple considerations imply that the magnitude limit which can be reached within about t 30 min at a given telescope and filter is a solid constraint which cannot be exp ∼ pushed by more than ∆mlim 0.5 mag with reasonable effort for large area survey projects. The last column of Tab. 7.2∼ states the galaxy dimming for the given filter from redshift 1 to 2. The rapid dimming in the optical R and I-bands of ∆m1→2 =3.4 compared to the moderate changes of ∆m1→2 =1.4 for H and Ks can be understood in terms of the K-correction introduced in Equ. 3.23. The long-wavelength filters H and Ks still probe the rest-frame peak emission of ellipticals even at very high redshifts, whereas the optical bands are restricted to the residual rest-frame UV-flux below the D4000 feature once the break is redshifted beyond the band. Consequently, reaching a depth to observe an m ∗ galaxy at z 1.5 in the R-band is already challenging even with the Very Large Telescope, but the task∼ becomes drastically easier when considering the longer wavelength-bands. The very same galaxy appears a factor of 100 brighter in flux in the Ks-band, roughly 60 times higher in the H-band, and the gain in the Z-band is still a factor of 7. For a high-redshift two-band imaging strategy, the choice of the I or Z-filter (instead of R) as the blue band hence results in a significant gain in terms of depth requirements at the faint end. From the galaxy properties point-of-view, the red filter selection should be at the longest feasible wavelength. However, as pointed out in Tab. 7.2 Ks-band observations are very challenging and time consuming (long observational overheads) due to the dominating thermal background contribution. The red filter of choice for a new approach is consequently the H-band, which combines (i) good high-z galaxy observability, (ii) high- redshift sensitivity in conjunction with I or Z, and (iii) improved background conditions compared to Ks. This discussion has shown that at the high-redshift end of the target regime around z 1.5 a modified choice of observation bands can (over-)compensate for the geometrical telescope∼ difference between the VLT and a 4 m-class telescope, which is of the order of 4–6 in collecting area. The next section will investigate the performance expectations concerning the redshift estimation of selected methods in more detail.

7.1.3 Performance expectations The standard R–Z method is now compared to the two alternative methods I–H and Z– H. For a quantitative performance benchmark of the redshift sensitivity some simplifying assumptions are needed. For the total error budget of a red-sequence measurement one has to consider (i) intrinsic scatter about the sequence, (ii) photometric measurement errors of the galaxies, and (iii) calibration errors of the photometric zero points. The intrinsic red-sequence scatter for all known clusters out to z 1.3 is typically σscatter < 0.05 mag. Of course, studying the evolution of the red-sequence∼ properties at the highest∼ redshifts will be an important science goal, but for now the assumption is that a well- deﬁned and tight cluster red-sequence exists and that the average color can be measured 7.1. A NEW NIR IMAGING STRATEGY 127

with good precision. Concerning the photometric measurement errors, the assumption of sufficiently deep observations in both filters has to be made. The discussion of the last section has shown (see Tab. 7.2) that this is not the case for the optical filters R, I, and Z. Depending on the telescope size, the limiting magnitude will unavoidably drop below m somewhere between redshift 1 and 2, and the bluer the band the earlier this will ∗ happen. The performance evaluation will thus assume idealized conditions, but in practice the achievable limiting magnitude of the blue filter will determine at what redshift the galaxy photometry will break down. Finally, calibration errors for the determination of the photometric zero points will add a constant color uncertainty. For well-calibrated data this offset should also be σZP <0.05 mag, which also neglects possible practical calibration issues. In any case, the total∼ achievable color error will have a constant part, due to zero point offsets, and a redshift dependent part which is related to the increasing photometric error of fainter galaxies, fewer objects on the red-sequence due to the magnitude limit and a possibly intrinsically less populated locus of galaxies. For the following discussion a total red-sequence color error of σ 0.05 (1 + z) mag is assumed, which is a reasonable color ≈ · parametrization for good quality data (σ 0.1 at z =1) out to z 1.5. color ∼ ∼ The expected redshift uncertainties σz can then be estimated from the first derivative of the color evolution as presented in Fig. 7.1 as

dz d(X Y) −1 σz σcolor = − σcolor , (7.2) ≈ d(X Y) · dz ! · − where (X Y) denotes the photometric method R–Z, I–H, or Z–H. The derivative of the − color evolution was computed numerically using a boxcar-smoothed input curve in order to suppress the influence of small scale features of the coarsely sampled model on the local slope estimation. The obtained redshift uncertainty estimates are shown in Fig. 7.2 for solar metallicity models and formation redshifts three and five. As expected, the best redshift performance is obtained in the redshift regime where the D4000 break is passing through the blue filter (see Tab. 7.1). The lowest achievable uncertainties of σz 0.04–0.08 are assumed in the redshift interval [0.3–0.9] for the R–Z method, [0.3–1.2] for≈ the I–H, and [1.0–1.4] for the Z–H approach. The main performance characteristics of different red-sequence methods are summarized in Tab. 7.3. Based on the former discussions and the performance expectations, the final evaluation of the different suitable red-sequence imaging strategies is as follows:

The R–Z approach has the advantage of being an optical method, i.e. observations • and data reduction tools are standard. However, sensitive Z-band instruments are rare, which basically limits an eﬃcient strategy to VLT FORS 2. The method has two systematic drawbacks for high-z cluster searches: (i) The redshift uncertainties at z> 1 increase to at least ∆z > 0.2 in the optimistic case (z = 5), and could ± f be as much 0.4, i.e. basically redshift-degenerate,∼ for a lower formation redshift model. Since± both bands have to be photometrically calibrated using standard star observations, there is little room to achieve a better performance by improving the calibration errors. (ii) The depth limitation in the R-band poses a serious risk to 128 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

not obtain sufficiently accurate color information on the galaxies or miss the highest redshift clusters completely. With a limiting magnitude of R 24.5–25.0 one can AB ∼ probe the cluster luminosity down to m*+1 at z =1, but by z =1.5 this has dropped to m*-1. If the cluster does not contain several galaxies brighter than the characteristic magnitude m*, the identification of a red-sequence could be an issue at redshifts as low as z 1.2. ∼ J–Ks has been the standard NIR approach for high-redshift cluster investigations • (Stanford et al., 1997). The color evolution (Fig. 7.1, central left panel) has a constant but shallow slope, with modest uncertainties at all redshifts. The advantage of the method is that sufficient depth in the J and Ks-bands to probe galaxies out to z 2 ∼ is fairly easy to achieve even with small telescopes. The disadvantage is the long overheads for NIR observations, in particular for Ks, which lowers the efficiency for routine follow-up. However, J–Ks can be very important to confirm very high-redshift cluster candidates that have been identified with any of the other approaches.

The I–H method shows very promising performance characteristics at all redshifts. • Sufficient I-band depth to at least m* at z = 1.5 should be achievable with 4 m- class telescopes with reasonable exposure times, which allows this approach to be implemented at many observatories. The disadvantage, however, is the need to use different instruments for the optical and NIR observations. This results in less flexibility during the observations, more scheduling constraints, and adds extra steps to the data analysis. This method is currently in the test phase with first data obtained with the EMMI and SOFI instruments at the New Technology Telescope (NTT) at La Silla.

The two-band imaging method of choice for this thesis is the Z–H strategy. This ap- • proach has the best performance expectations in the target redshift regime 1

Color Instrument σz in 0.51.5 Comment ∼ ∼ R–Z optical 0.05–0.15 0.15–0.4 no sensitive to zf I–H optical+NIR 0.06 0.06–0.15 possibly instrument change Z–H NIR (+opt.) 0.08–0.12 0.06–0.1 yes J–Ks NIR 0.08–0.14 0.14–0.18 yes long overheads

Table 7.3: Performance summary of the diﬀerent imaging methods. Columns 3 and 4 state the estimated redshift errors as discussed in Fig. 7.2. Column 5 indicates whether the method has the potential to be used at redshifts z>1.5.

Figure 7.2: Absolute redshift uncertainties of diﬀerent red-sequence methods for a realistic photometric color error assumption of σ 0.05 (1 + z) mag. Error estimates were obtained color ≈ · from the derivatives of the smoothed model colors in Fig. 7.1 using σ dz/d(X Y) σ , where z ≈ − · color (X Y) denotes the photometric method. The black solid line illustrates the estimated redshift error − for the R–Z technique under the assumption of a formation redshift of zf = 5, blue shows the Z–H method, and red I–H. The dotted lines use a model formation redshift of zf =3 for the same methods. The R–Z color is an accurate redshift estimator up to about z 0.9, but at z > 1 the uncertainties ∼ grow drastically and for low formation redshift models, the z-estimates even become∼ degenerate (ﬂat slope in lower left panel of Fig. 7.1). The Z–H strategy (blue) yields good redshift estimates all the way to z 1.5, has the lowest uncertainties in the targeted regime 1

Figure 7.3: Left: The NIR wide-field camera OMEGA2000 at the Calar Alto 3.5m telescope. Right: OMEGA2000 instrument response. The green solid line indicates the transmission efficiency through the optical system as a function of wavelength, the blue line shows the HAWAII-2 detector quantum efficiency, and the black solid line gives the full response of the camera system. Filter band boundaries are overplotted as vertical lines. With a quantum efficiency of > 70%, the Z-band sensitivity of the OMEGA2000 NIR array is comparable to or better than the best∼ optical instruments in this critical filter band. Data from H.-J. Röser (private communication).

7.2 Practical Implementation

The Z–H two-band imaging strategy can be ideally carried out with the NIR wide-ﬁeld camera OMEGA2000 at the Calar Alto 3.5 m telescope. This section will brieﬂy introduce the capabilities of the instrument, the observing strategy, and the obtained data.

7.2.1 OMEGA2000 in a nutshell The left panel of Fig. 7.3 shows the prime focus instrument OMEGA2000 mounted at the 3.5 m telescope. Three main features establish this camera as the instrument of choice for testing the Z–H approach: (i) the high sensitivity from 0.8–2.5 µm, (ii) the large 15.′4 × 15.′4 field-of-view, and (iii) an online reduction system for an immediate data evaluation. The instrument sensitivity is shown in the right panel of Fig. 7.3 as a function of wavelength. The blue solid line illustrates the quantum efficiency (QE) of the HAWAII-2 NIR array, the green line follows the transmission properties of the optical system of the camera, and the black line shows the total response of OMEGA2000 as the product of the two former functions. The QE efficiency in the classical NIR band J, H, K is about 80%, and at the short wavelength end in the Z-band one expects still a 70% sensitivity. For comparison, the red-optimized CCDs of the VLT FORS 29 have an average Z-band QE

9See http://www.eso.org/projects/odt/Fors2/qeu.html. 7.2. PRACTICAL IMPLEMENTATION 131

Instrument Detector field-of-view 15.′4 15.′4 (237 arcmin2) array type HAWAII-2 × pixel scale 0.45′′/ pixel array size 2048 2 048 pixels weight 300kg pixel size ×18 µm focal station prime focus wavelength range 0.8–2.5 µm focal ratio f/2.35 quantum efficiency 70–85% telescope primary 3.5 m dark current < 0.03 e− s−1 broad band filters Z, Y, J, H, Ks, K read noise < 15 e− data rate 10GB / night good pixels >99% online reduction ∼ yes fill factor 90% commissioned 2003 full well capacity 100000e− ≃ Table 7.4: OMEGA2000 instrument and detector characteristics. From Röser (2005) and Haas (2002).

of 65%, but most optical instruments provide quantum efficiencies <30% in Z. Modern NIR arrays hence offer an attractive sensitive alternative at the long-wavelength end of the traditional optical regime. An additional advantage of NIR arrays with their individually isolated pixels is the absence of fringing, i.e. interference patterns caused by strong monochromatic atmospheric emission lines in CCDs; the disadvantage on the other hand are the numerous imperfections and flaws of NIR detectors that have to be accounted for during the reduction process (see Sect. 7.3). The second prime feature of OMEGA2000 is the large field-of-view, which is still one of the largest available fields in the near-infrared. The 237 square arcminute FoV covers one quarter of the XMM-Newton field and is five times as large as the FORS 2 FoV. This has several important advantages: (i) often several cluster candidates can be simultaneously observed, (ii) lower redshift candidates and calibration sources can be placed in the FoV at no extra observational cost, and (iii) the photometric calibration in the J, H, and Ks bands is simplified by taking advantage of the numerous 2MASS sources in the field (see Sect. 7.4.2). The available OMEAG2000 online reduction system is an additional important characteristic which helps to optimize the observational efficiency. Since the NIR background is so dominant (see Tab. 7.2), an evaluation of the data quality and the depth of the observation is usually difficult to obtain, in particular if the targets of interest are very faint. The online reduction pipeline can provide quicklook data products with sufficient quality while the data taking is active. This way, the reduced and co-added images of the ongoing observation are available with an approximate delay-time of 3 min to the running exposure and thus allows an instant evaluation of the target source and the observational status. Further OMEGA2000 instrument and detector characteristics are summarized in Tab. 7.4. 132 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

7.2.2 Observational strategy As a follow-up program of a large sample of X-ray selected cluster candidates, the observational survey mode strategy of the Z–H approach is driven by the two main goals (i) source identification and (ii) efficiency. The first goal implies that the observations have to be deep enough in order to identify even the most distant clusters as local galaxy overdensities. On the other hand, the bulk of clusters at lower redshifts require much shorter exposure times for a safe identification, which increases the survey efficiency if properly accounted for. For these considerations the following observing strategy has been pursued:

1. Start H-band observation of a distant cluster candidate ﬁeld;

2. Check the online reduced co-added image products every 4–5 minutes and evaluate the appearance of the cluster candidate; 3. Stop the H-band observation once a galaxy cluster signature is clearly visible with at least 10–15 galaxies close to the candidate X-ray center; 4. If the target shows no cluster signature, stop after an exposure time of 50 min, which corresponds to a limiting 5 σ H Vega magnitude of H >21 (m*+1 at z 1.5). The lim ∼ candidate source is classified as ‘spurious’ and is not obser∼ ved in any other band; 5. Observe sources with cluster signatures in the Z-band. Check again the pipeline products for sufficient depth and abort observations once reached; 6. Go to next candidate target field; 7. Obtain J and Ks-band observations for some of the most distant candidates to backup with a J–Ks color.

Starting with the redder H-band is important for the identification of the highest redshift clusters, which might be heavily dimmed in the Z-band (see Tab. 7.2), and correspondingly for the secure classification of a null-detection, i.e. spurious candidates. Galaxy clusters at redshift z < 1 can typically be well identified with exposure times of texp <15–20 minutes, i.e. the efficiency∼ gain compared to the deepest observations is about a factor∼ of three. The photometric Z-band calibration requires additional observations of SDSS Z-band standard stars several times per night. Each targeted Z-band field also needs at least a few minutes of coverage under photometric conditions, which might have to be re-observed if the conditions during the initial exposure were not sufficient in quality. The total exposure time texp for a deep observation is split up into many shorter observations ttel, after which the telescope is offset, and these again consist of the sum of several single frames of integration time tsingle. The elementary exposure time tsingle is motivated by two boundary conditions. Firstly, the background in a given filter sets the maximum time before the detector saturates. Secondly, the difference between the saturation limit and the background level defines the effective dynamic range of the exposure, i.e. the interval in which the flux of astronomical targets can be accurately measured. In practice 7.2. PRACTICAL IMPLEMENTATION 133

this time is set in a way to include the brighter calibration objects well within the dynamic range and was fixed at tsingle = 5 s (10 s) in the H-band (Z-band). These single exposures can be co-added in memory before they are saved to disk after a time ttel during which the telescope points at the same position. The maximum time for ttel is limited by the timescale of the temporal sky background variations which is a few minutes in the NIR and is less critical in bands with a lower background level. ttel, as the exposure time of individual saved frames, is then directly linked to the data reduction strategy for the sky modelling (Sect. 7.3) and is set to ttel =40 s (60 s) in the H-band (Z-band). After the time ′′ ′′ ttel the telescope is dithered, i.e. moved by a small pre-defined offset of 20 –30 in order to place the objects at a slightly different position on the detector for the next exposure. This procedure is repeated until the total exposure time texp = n ttel = n m tsingle, with n the number of telescope positions and m the number of in-memory· co-adds,· · is sufficient to reach the required depth. The total overheads10, i.e. the time the camera is not observing the target, can be 50–80% of the actual integration time and thus amount to a significant time fraction to be carefully considered and optimized within the requirements of the scientific program.

7.2.3 Imaging data overview Following the above observing strategy, we have conducted two runs at the Calar Alto observatory, which were complemented by subsequent service observations. Table 7.5 summarizes the observation runs and the data yields. The total observing project was designed as a 10-day program to follow-up XDCP candidates in the Northern part of the survey region at 20◦

Run Date Good Nights Data Quality Raw Data Included Calar Alto I Jan 06 2.5 / 10 ok / non-phot. 29GB yes CA Service A May 06 0.2 fair / partly phot. 4 GB yes Calar Alto II Nov 06 3.5 / 13 good / photometric 40GB yes CA Service B Jan 07 1.0 fair / partly phot. 12GB yes CTIO I Sep 06 2.0 / 3 ok / non-phot. 27GB no

Table 7.5: NIR data overview. The data for this thesis was obtained during two Calar Alto runs in 2006 complemented by two service observation blocks. Including data that was taken in challenging conditions, the proposed observing program achieved a satisfactory completion level of about 80%. The results of the CTIO run in the last row using the ISPI instrument are not included in this thesis, since the data is not self-contained (H-band only) and complements an ongoing optical imaging program.

7.3 Development of a NIR Science Reduction Pipeline

Near-infrared wide-field imaging instruments have only emerged over the last few years and OMEGA2000 in this respect is still a camera with one of the widest available field-of-views. Additionally, the NIR data reduction process is more challenging and complex than in the optical. Both aspects combined result in the fact that there is no generic ready-to-go data reduction software package available yet that would fulfill all demands. This gave rise to the motivation to develop a new science-grade NIR reduction pipeline optimized for the XDCP demands.

7.3.1 Requirements and specifications From the expected science target characteristics discussed in Sect. 7.2.2 and the raw data properties of the last section, we can now define the requirements and demands on a science- grade reduction pipeline. The distant galaxy cluster detection and characterization result in the scientific requirements of:

the best possible performance for faint extended objects; • the maximum achievable depth for the detection of the faintest objects; • accurate object photometry down to the limiting magnitude; • quality control via intermediate data products. • Practical demands on the other hand motivate the following pipeline speciﬁcations of:

handling of thousands of raw images from dozens of pointings and several ﬁlter bands; • minimal requirements of manual interaction; • 7.3. DEVELOPMENT OF A NIR SCIENCE REDUCTION PIPELINE 135

handling of strongly varying observing conditions and data contamination; • an OMEGA2000 optimization with the ﬂexibility to easily adjust to other NIR in- • struments; a portable software package for easy public distribution and use for similar survey • projects. The science-grade NIR pipeline developments built upon the existing quicklook reduction pipeline which is discussed and characterized in detail in Fassbender (2003). The following discussion will mainly focus on the upgraded pipeline components and will only summarize the functionality of existing critical components. The reduction pipeline is implemented as a C-application program within the MIDAS environment. The MIDAS system provides the interface and tools for the image data handling, whereas all pixel-by-pixel image manipulations are based on C-coded algorithms.

7.3.2 NIR reduction steps The implemented near-infrared data reduction scheme is illustrated in Fig.7.4. Data products for the different reductions steps from the raw image to the final science frame are provided in Fig. 7.5. The full data reduction procedure can be broken up into the two independent processing blocks (i) single image reduction and (ii) image summation. The pipeline is designed in a way to yield final image products for one or both of the processing blocks without further manual interaction. The user provides an input catalog with an arbitrary number of raw images belonging to the same telescope pointing and taken with the same filter and upon pipeline start all further image handling procedures are automated.

Single image reduction The first processing block handles the calibration steps of the individual raw frames (upper left panel of Fig. 7.5 and subpanel 1) and the subsequent sky modelling and subtraction as shown in Fig. 7.4. As initial calibration step, the raw input image is flatfielded to correct the detector inhomogeneities. For NIR detector arrays this is particularly critical since (i) the quantum efficiency can vary by a factor of two over the FoV (global variations), and (ii) the electroni- cally independent NIR pixels exhibit significant pixel-to-pixel changes (local variations). A master flatfield for each filter can be created from dome flatfields taken inside the telescope dome and sky flatfields taken during twilight in a region devoid of bright objects. Since the thermal dome emission imprints a significant amount of structure in the dome images, the master flatfield mainly relies on the detector variations determined from the sky calibration data with the dome flatfields acting only as initial correction of the former. The final master flatfield contains the information on the global and local detector variations. It is normalized to a median value of 1 and placed in a calibration frame directory, where the reduction pipeline can access it. By dividing the raw input frame by this master flatfield (subpanel 2) the intermediate image product contains variations due to the received signal and not the detector. 136 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

Figure 7.4: Near-infrared science reduction pipeline ﬂow chart. 7.3. DEVELOPMENT OF A NIR SCIENCE REDUCTION PIPELINE 137

Figure 7.5: Near-infrared science reduction steps. The top panel shows the full 15.4′ 15.4′ FoV × for a raw H-band image (left) and the final reduced sum image containing 75 input frames. Note that about 30′′are lost at each field edge due to the dithering (right). The blue squares indicate the 2′ 2′ zoom field for the small panels centered on the galaxy cluster RX J0018.2+1617 at z = 0.55. × The small cutouts with numbers 1-12 illustrate the reduction steps of Fig. 7.4 by showing intermediate data products. 138 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

As a second step, a bad pixel correction is applied. The OMEGA2000 detector array contains about 0.5% pixels which are either not functioning at all (dead pixels) or exhibit significant deviations from a time-linear signal response (hot or flickering pixels). These pixels can be identified through a linearity analysis and are flagged in the bad pixel mask (subpanel 3) of the calibration directory. The values of these flagged bad pixels in the flatfielded image are then replaced by the interpolated pixel value of the four closest good pixels. Note that near-infrared measurements are usually obtained by reading out the detector array at the beginning and end of the integration time and subtracting the two to determine the signal. Hence this double-correlated readout scheme does not contain a bias offset as in CCD data. The detector dark current, i.e. the measured pixel values without an external signal, is in principle an additional (small) contribution to be corrected. However, tests have shown (H.-J. Röser, private communication) that the dark current is not stable enough to be accurately modelled and subtracted since it depends sensitively on the exact detector temperature and its thermal history. Instead of an explicit subtraction, the dark current is treated as part of the overall background and implicitly corrected with the sky background subtraction. The most critical and important part in the near-infrared reduction process is the sky background modelling. The dominating background components are (i) atmospheric airglow (see Fig. 8.1) and (ii) thermal emission of the ambient structures at long wavelengths. Smaller contributions arise from (iii) scattered moonlight, (iv) Zodiacal scattered light (Cox, 2000), and (v) the uncorrected instrumental dark current. As discussed in Sect. 7.1, the science objects of interest can be easily a factor of 100–1 000 fainter than the background surface brightness. Consequently this implies that the local background around the astronomical objects has to be known with uncertainties of < 10−3. The dominating airglow emission exhibits significant temporal changes on time scales of several minutes and spatial variations on scales of a few arcminutes. Hence the sky modelling requires a temporally and spatially local background estimation derived directly from science data. The pre-requisites to an accurate sky modelling have been implemented in the observing strategy through (i) a sufficiently short telescope pointing time ttel that allows tracing the temporal variations and (ii) well-chosen dither offsets that enable a reconstruction of the local background without being biased by the flux of the astronomical objects of interest. The following recipe yields an accurate but preliminary estimate of the sky background. (i) Take the image together with the three11 preceding and following frames, i.e. the background is determined from a total of seven frames taken within a period of 3–5 minutes. (ii) For every detector pixel (image coordinates) select the corresponding pixel± values of all seven frames. Due to the dither strategy, which places the real objects at different detector positions for each image, most pixel values will only contain a background signal, whereas a minority will additionally have an object signal component12. (iii) The median

11Optimized pipeline parameters for the XDCP NIR data reduction are cited for this discussion. How- ever, all critical parameters of the software can be specified and adjusted to the actual science needs. The background for the first (last) images of an observation series are estimated from the first (last) 7 frames. 12This assumption is in general well justified for extragalactic sparse fields. However, for (low redshift) 7.3. DEVELOPMENT OF A NIR SCIENCE REDUCTION PIPELINE 139

of these seven values is taken as a good initial guess of the actual background level for the specified pixel. However, any object flux in one or more of the seven values systematically biases the median to a slightly higher level, i.e. the background is overestimated due to the additional non-background component in some pixels. (iv) This background bias can be corrected in next order by applying a κσ-clipping algorithm to the data values. A Poisson estimate of the local background standard deviation σ is obtained from the initial background estimate, the gain13 of the detector array, and the local quantum efficiency of the detector pixel14. Any pixel values deviating more than three standard deviations from the initial background estimate are clipped, and the median is recomputed from the remaining object-cleaned values. (v) The κσ-clipped median value is taken as the background estimate for the specified pixel. The procedure is repeated for all detector pixels resulting in a full sky frame for the input science image. (vi) The modelled background sky is subtracted from the science image and replaced by a constant representing the average sky value in order to preserve the proper counting statistics. The flatfielded, bad pixel corrected, and sky subtracted frame is written to disk as the final result of the first single image reduction loop (subpanel 4). However, the applied κσ-clipping for the first order correction of the local sky value could only eliminate object flux which has been detected at more than three standard deviations above the normal background level in the single frame, i.e. for fairly bright sources. The bulk of fainter sources and the wings of brighter objects below this clipping threshold will still bias the local background model to slightly higher values. The subtracted sky might thus still be overestimated resulting in a systematically lower flux for an object at the given position, which can reach significant levels for faint galaxies (see Sect. 7.3.3). This lost flux can be largely recovered with a second iterated background modelling loop as discussed below.

Image summation

The reduced single image contains the data of integration time ttel, i.e. typically 60 seconds or less. In order to reach the required depth, the single images have to be stacked to build up a deep sum image with the full on-target exposure time texp. To achieve this, the dither offsets have to be reversed and the images aligned in world coordinates, i.e. the physically same objects in the individual frames have to end up on top of each other. During this stacking procedure cosmic ray events, or short cosmics, can be identified and removed. The alignment and stacking procedure uses the following approach: (i) The first input image is declared as master frame onto which all others are positioned and co-added. (ii) The reduced single master image is searched for typically 50–100 reference stars, whose centers-of-mass, as derived from the stellar flux distribution, are recorded. (iii) From the

objects larger than the dither offsets of 20′′–30′′ the assumption is not fulfilled, but the introduced initial error is corrected during the second iteration. 13The gain of a detector is the conversion factor between digital count units and physically detected electrons via the photoelectric effect. 14In practice, an additional normalization factor has to be considered which is necessary to standardize the background levels in each frame. 140 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

image header of the next input frame, the expected offsets with respect to the reference master frame are computed. Using these approximate offsets, the catalogued reference stars are searched for in the vicinity of the expected detector positions. Once matched, the exact object centers are determined resulting in an improved offset determination. (iv) The final image offset is determined from the outlier-clipped average of all matched stars. Since the OMEGA2000 optics exhibits negligible image distortions over the whole FoV, the frames can be co-added at this stage by simple XY-coordinate shifting to the nearest integer pixel offsets. The single images still contain the unwanted cosmics which are to be corrected for without modifying the flux of the real astronomical objects. Cosmic ray events, caused by charged particles going through the detector, can be identified from their characteristic signature of single or few isolated pixels with increased count levels without a corresponding counterpart in the next aligned image. In principle, the cosmics would be removed by a simple median process over a sufficient number of pixel-aligned images, similar to the procedure for the background determination. The complication arises that the median procedure efficiently removes everything that is not present in at least half of the images, including the outer wings of real astronomical objects in variable seeing conditions, and the central flux peak of objects. The peaks can exhibit large gradients towards the neigh- boring pixels and are thus likely damped when medianing over the different object flux distributions in each frame. This effect can decrease the total flux of stellar sources by about 20%, which is avoided by applying the following procedure: (v) The median image of eleven world-coordinate aligned input frames is determined and used as object reference image. This intermediate sum image is now devoid of cosmics and contains the real objects at the proper locations but with biased flux levels. (vi) Each individual frame is now checked pixel-by-pixel whether the value is more than five standard deviations above the corresponding median reference value, where the standard deviation is determined in analogy to the sky modelling procedure. (vii) If a cosmics candidate is detected, the reference median image is searched for an astronomical object within a radius of three pixels around the outlier position. In case a source is found, the initial pixel value is classified as object flux and co-added to the master image as is. If no source is matched, the outlier value is classified as cosmic and replaced by the median value of the reference image, which is in turn co-added to the master sum image. (viii) This procedure is repeated until all input images have been aligned, cosmics cleaned, and co-added to yield a final deep sum image (subpanel 5). The cosmic ray events, as shown in subpanel 10 of Fig. 7.5, have now been removed from the field without influencing the total flux of the astronomical sources. The last two subsections have provided a summary of the full first data reduction loop, which is in principle equivalent to the quicklook reduction system available at the telescope. The next two subsections will discuss further improvements towards the goal of the best achievable science-grade data reduction quality.

Iterated background subtraction The complete science-grade reduction process requires two additional main components. Firstly, the creation of an object mask with registered positions of all sources. Secondly, an additional iterated loop through the reduction procedure but now with removed source ﬂuxes for an unbiased second order sky background correction. 7.3. DEVELOPMENT OF A NIR SCIENCE REDUCTION PIPELINE 141

The object mask is created from the final deep sum image of the first reduction pass (subpanel 5) which contains the best available information on the location and size of the astronomical objects at this processing stage. The objective of the object mask is to cover all sky areas in the FoV which contain a detectable source signal and use only the pixels with an unbiased background value for the reconstruction of the sky model. In practice the mask is created using an iterative κσ-clipping algorithm applied to the smoothed deep sum image. An initial smoothing is helpful for the identification of faint object halos, i.e. the PSF wings for point sources are extended halos for galaxies. Pixel values deviating more than about two standard deviations from the median background level are classified as object pixels and flagged in the mask frame. This procedure is iterated to push the depth of the mask to sufficiently faint levels. The final object mask is shown in subpanel 6, which now distinguishes the black regions with background pixel values and the masked white areas with object flux contributions. With the object mask in hand, the second reduction loop can start from the very beginning (see Fig. 7.4). The initial calibration steps flatfielding and bad pixel correction are applied as before. The sky modelling procedure on the other hand is extended by two additional steps. (i) The object mask is projected onto each individual input frame. To do this projection properly, the exact image dithering offsets as determined during the first image summation loop are used to shift the object mask to the correct world-coordinate aligned location in each input image. (ii) The pixel values of the masked object regions are replaced15 by the unbiased median background level of the 100 nearest unmasked values. This replacement scheme is designed to yield robust results for arbitrary large objects and to take into account local background variations across the FoV. The masked input image (subpanel 7) with all object flux replaced is then passed on to the sky modelling procedure. The final unbiased sky model (subpanel 8) is now the result of a second order corrected flux- removed and κσ-clipped background determination. Subtraction from the original input image yields the final iterated reduced single image (subpanel 9), and after the subsequent summation and cosmics removal (subpanel 10) of all frames the final co-added deep sum image is available after the second processing loop (subpanel 11).

Advanced optimizations The final sum image is a well-reduced and science-ready data product. However, with respect to the objective of reaching the maximum possible depth, two additional optimization schemes can be applied. Firstly, fractional pixel offsets instead of integer values can be implemented for the stacking procedure, and secondly, optimal weighting of the individual input frames can improve the final signal-to-noise ratio. The discussed image summation procedure determined the accurate XY-offsets of the frames with respect to the master frame, but then rounded the exact values to the nearest integer pixel offsets in order to match the underlying pixel grid. This way, systematic

15Two additional replacement schemes with lower performance expectations are implemented. 1) The masked pixel is ignored for the sky determination, which leads to diﬃculties for very large objects. 2) The masked pixel is replaced by the global median, which does not take into account local variations. 142 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

truncation errors of up to half a pixel in each spatial direction are unavoidable. For the OMEGA2000 pixel scale of 0.45′′/pixel this implies systematic XY-shifting errors of 0–0.225′′ per direction, which can add up in quadrature to a maximum systematic dis- placement error of 0.32′′ for a stacked frame. The integer pixel shifting has thus two consequences, both gaining importance in good seeing conditions. (i) The stacked object flux is not as compact as defined by the seeing limit and (ii) the 2-dimensional flux distribution of point sources does not follow its natural radially symmetric Gaussian shape. This latter tendency of the object cores to become boxy or square-like is easily visible for the fainter objects in the left panels of Fig. 7.6. The shape improvements after applying the fractional pixel offset scheme can be seen by eye in the right panels of the same figure. In order to implement fractional offsets onto the underlying fixed detector grid of 2048 2 048 pixels, the flux values in each pixel have to be distributed over four cells of the× master grid in a flux conserved way. The flux distribution matrix is calculated once for each frame from the exact offsets according to the fractional geometrical overlap of a pixel with the underlying master grid. The pixel values of the frame to be co-added are then distributed using this pre-determined matrix. In turn, each pixel of the master grid, receives fractional flux contributions from four adjacent pixel values of the single image. The question whether a faint object will be detected depends primarily on the signal- to-noise ratio of the source (see Equ. 7.1), i.e. on its contrast with respect to the noise properties of the background. So far, all input frames for the summation were treated as if they had the same SNR properties. However, this is in general not the case since the observing conditions for the seeing σ, the sky transparency T , and the background surface brightness levels B can change significantly during the observations. The source signal s will then vary as s T and the total background b confined in a seeing limited aperture will follow b B σ∝2. One can then ask the question for the best way to co-add two individual frames∝ in order· to achieve the best SNR for their sum. Following Gabasch (2004) for the case of faint sources (b s), an optimal weight factor w2 for the second ≫ 16 2 −1/2 frame can be derived by maximizing the resulting SNRtot =(s1 + w2s2) (b1 + w2 b2) . Using dimensionless parameters for the source signals ˜ s /s , the transparency· · change ≡ 2 1 T˜ T2/T1, the background variation B˜ B2/B1 or ˜b b2/b1, and the seeing ratioσ ˜ σ2/σ1 of≡ the two observations, the optimal weight≡ for the second≡ frame is given by ≡

s˜ T˜ f˜ w2 = = = . (7.3) ˜b B˜ σ˜2 B˜ σ˜2 · · In the last step, the change in the sky transparency T˜ was replaced by the equivalent but directly observable source ﬂux ratio of the two observations f˜ f /f = T˜. The optimal ≡ 2 1 weighting factor w2 for a new image to be co-added to the master frame is thus proportional to the received signal ratios ˜ and inversely proportional to the total background ratio ˜b in the object aperture, which is particularly sensitive to seeing variations. This is expressed in terms of the direct observables f˜, B˜, andσ ˜ that allow an automatic monitoring for each

16 2 2 2 2 For the second factor, the background noise is added in quadrature σtot = σ1 +(w2σ2) = b1 +w2b2 . Note that in this relation σ denotes the standard deviation and not the seeing. 7.3. DEVELOPMENT OF A NIR SCIENCE REDUCTION PIPELINE 143

image. The background factor B˜ is determined from the median sky level in each frame. The flux ratio f˜ and the seeing ratioσ ˜ can be obtained by averaging the stellar source properties as measured in the first step of the summation process, where all matched objects are characterized in terms of their 5′′-aperture flux and the full-width-at-half-maximum (FWHM) of their radial source profiles. In practice, the fractional pixel offset and optimal weighting schemes are combined and implemented as a separate optimized final sum image, as shown in the last subpanel 12 of Fig. 7.5. The effects of these additional optimization procedures as part of the science-grade data reduction will be investigated in the next section 7.3.3. This last section has discussed the main steps and procedures to go from raw NIR data to science-ready optimized stacked images and how they are implemented as part of the newly developed NIR reduction pipeline. For the routine reduction of the data listed in Tab. 7.5 the following approach was pursued for each science field and filter band: (i) A full automatic pipeline reduction is run on the field with standard parameters. (ii) The data quality of the intermediate and final data products is inspected. This includes checks of the individual reduced images for possible data corruption, contamination from scattered light, cloud coverage, and satellite trails. Additionally, the object mask is investigated for proper depth, the masked input images for a correct mask projection, and the sky models and cosmics frame for unusual features. (iii) Identified contaminated data frames are then excluded from the reduction and the pipeline parameters are optimized if necessary. (iv) The reduction pipeline is then re-run on the optimized input data set to yield the final science-grade image products.

7.3.3 Faint object optimization After introducing the basic functionality of the main science pipeline components in the last section, we will now turn to a discussion of the effects of the implemented reduction optimizations. Figure 7.6 illustrates the different optimization schemes for 1′ 1′ co-added image cutouts centered on the calibration cluster RX J0018.2+1617 at z =0.×55, displayed with high image contrast. The upper panels show the sum images after the first reduction loop, whereas the lower panels have the results after the iterated background subtraction. Going from left to right adds the optimal weighting scheme and the fractional pixel offsets. Hence the upper left panel displays the reduction quality achievable with a single loop (quicklook) reduction, and the lower right panel shows the same data with all optimizations incorporated requiring about the threefold CPU time. We will first consider the effects of the iterated background subtraction, i.e. the difference between the upper and lower panels of Fig. 7.6. By masking out all objects and their halos during the second reduction loop, the sky background determination is unbiased compared to the systematically slightly higher median levels when source fluxes are included. This lower background value at the object positions results in additional object flux when the sky background is subtracted. The iterated background subtraction hence recovers object flux and consequently makes the sources brighter. This flux recovery is expected to be 144 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

Figure 7.6: Faint object optimization. The upper left panel shows a high-contrast 1′ 1′ zoom on × the calibration cluster RX J0018.2+1617 at z = 0.55 as seen in the stacked sum image after the first reduction pass (see panel 5 in Fig.7.5). Going to the right panels adds fractional pixel offsets and optimal weighting. The lower panels show the image results after the second iterated sky subtraction. The final best image including all optimizations is thus placed in the lower right panel. While the iterated sky subtraction recovers a significant amount of flux of faint and extended objects, the optimal weighting scheme increases the limiting magnitude by improving the signal-to-noise ratio of the faint sources (see Fig. 7.7 for quantitative results). more significant for extended low surface brightness objects compared to stellar point sources. A useful approximation to convert measured magnitude offsets into flux differences ∆f =f f for small variations (∆m 1) is given by the relationship 2 − 1 ≪

f ∆f 2.5 ∆f ∆f ∆m = 2.5 log 2 = 2.5 log 1+ 1.1 , (7.4) − · f1 ! − · f1 ! ≈ −ln(10) · f1 ≈ − · f1 implying that small magnitude differences reflect approximately the fractional flux difference. 7.3. DEVELOPMENT OF A NIR SCIENCE REDUCTION PIPELINE 145

Figure 7.7: Effects of advanced reduction schemes. Top panels: Difference photometry in the H-band for various reduction stages as shown in Fig. 7.6 using the same deep detection image. Left: Total magnitude difference between the sum image with iterated background subtraction (here H1, lower left panel in Fig.7.6) and the sum image after a single reduction loop (here H2, top left panel in Fig. 7.6). The dashed red line represents a zero offset (i.e. the same measured magnitudes), black crosses show all objects in the FoV, big blue circles represent the galaxies seen in Fig.7.6 (i.e. cluster members of RX J0018.2+1617), and the green solid lines follows the median offset taken in 0.5 mag bins. At H>16, the recovered flux from the iterated background subtraction becomes noticeable and increases to a median difference of 0.1–0.15 mag (about 10-15% in flux) between H magnitudes of 18–21. At H>21 the difference drops back to zero which reflects the depth of the object mask. Note that the recovered flux is drastically increased for the faint extended cluster galaxies (blue circles) compared to the median difference. Right: Same plot for the difference between the final weighted image (H1, lower right panel in Fig. 7.6) and the iterated background subtracted image without weighting (H2). The total magnitudes are now consistent, since weighting cannot change the flux, but only the signal-to-noise ratio of the sources. Left bottom panel: log N–log S of the final reduced image (black solid line) and the first sum image (red solid line) in a field with significant seeing and transparency variations during the observations. The systematic vertical offset is due to the iterated background subtraction. The difference in total number of detected sources on the other hand can be attributed to the optimal weighting scheme. The effect is separated in the right panel, which shows the results of the best final image (black) and the non-weighted image (red). The limiting magnitude is significantly improved by the weighting and the total number of detected sources can be boosted by as much as 30–40% in fields with strongly varying conditions. 146 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

The effect of the iterated background subtraction is quantified in the upper left panel of Fig. 7.7 where the difference photometry with and without the second reduction loop is shown against the total object magnitudes. For both images the measurements were performed in the exact same apertures by using a common deep multi-band detection image (see Sect. 7.4.1), i.e. any offsets are due to object flux differences. No measurable effects would thus imply zero offsets (dashed null line), and more object flux after the iterated loop will manifest itself in a positive magnitude offset (the brighter, i.e. smaller, magnitudes are subtracted). The measurements for all objects are plotted as black dots, for the cluster members in Fig. 7.6 as blue circles, and the median averaged differences in 0.5 mag bins are traced by the green solid line. Following this line shows practically matching magnitudes out to H 15, which is expected due to the photometric calibration procedure using bright 2MASS∼ stars as discussed in Sect. 7.4.2. At H>16, the recovered flux from the iterated background subtraction becomes noticeable, i.e. the magnitudes are systematically brighter, which increases to a median difference of 0.1–0.15mag (about 10– 15% in flux) between H magnitudes of 18–21. At H>21 the averaged difference drops back to zero reflecting the depth of the object mask which does not cover the faintest sources anymore. Note that the recovered flux is drastically increased for the cluster galaxies (blue circles) compared to the median difference. The difference for the bright cD galaxy with its extended halo is 0.1 mag and in the important regime between 18

more than one standard deviation above the background (see Sect. 7.4.1). However, for a stacked sum of only ten images (typically 20–80) this would already require a single image pixel-noise outlier of σsingle > 4√10 12 above the background, which implies that it would have been easily detected∼ and removed≈ by the cosmics removal procedure. Even a constantly hot pixel not included in the bad pixel mask would most likely not lead to a spurious source since (i) it would show up in the sky model and be subtracted with it, and (ii) these outliers would again be removed from the cosmics routine since they do not align in world coordinates. In conclusion, the fractional pixel offsets procedure is not expected to cause any significant number of spurious source detections through correlated noise. The last step towards the maximally achievable image depth is the implemented optimal weighting scheme. The measurable effects of this optimization strongly depend on the variability of the observing conditions according to Equ. 7.3. However, since the execution time per field and filter is typically 30–80 min long, changes of the external conditions are likely, leading to significant improvements if weighting is applied. Variations that modulate the weighting factors include (i) pronounced background changes at the beginning or end of the night or due to the moon setting or rising, (ii) transparency drops induced by partial cloud coverage, and most sensitively (iii) seeing variations in the atmosphere or the telescope dome, e.g. if the primary mirror is warmer than the ambient air. The optimization of the signal-to-noise ratio at the faint source end has two important effects. Firstly, objects can be characterized with higher photometric accuracy, i.e. the measurement errors decrease. Secondly, very faint formerly undetected sources are now pushed over the detection limit requiring a minimum fixed SNR. This latter effect is of particular importance since the total number of sources increases rapidly with the flux limit as N f −3/2 (Equ. 3.26) and allows to probe the galaxy luminosity function in tot ∝ lim clusters to fainter levels. The number counts for a field observed under strongly varying conditions in the lower right panel of Fig. 7.7 illustrates the achievable gain. The black curve shows the detection results for the image with the applied optimal weighting scheme, the red line gives the outcome on the non-weighted image. Up to H<20 the number counts are basically identical but at H 20.5 the red non-weighted curve∼ turns over and reaches ∼ its plateau. The black weighted line on the other hand exhibits a significantly improved limiting magnitude, which in turn boosts the total number of detected sources for this field by almost 40% at no extra observational cost. The depth optimization via fractional pixel offsets and optimal weighting thus takes effect in the magnitude range 2016 due to the discussed iterated background subtraction, which is the consequence of the objects being systematically brighter, hence more sources exist at a given fixed magnitude. In summary, the implemented optimization schemes17 for the science-grade reduction

17The faint object optimization eﬀects in this section have been discussed using H-band data, since this is the master ﬁlter for the Z–H imaging strategy. The achievable improvements scale with the background 148 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

pipeline have an important impact on the targeted science applications. The iterated background subtraction corrects the systematic biasing of the total object magnitudes, whereas the combined effects of the fractional pixel offsets and the optimal weighting scheme improve the limiting depth of the data. With respect to the intended distant galaxy cluster identifications the achieved improvements are of prime interest. (i) The measured object magnitudes need to be accurate in the full range 16

7.3.4 Pipeline and instrument performance In this section we will apply the NIR reduction pipeline to real multi-band data of the XDCP follow-up program, compare the achieved depth to the requirements of Sect. 7.1.2, and discuss the expected background object densities. The analysis is performed in the follow-up field North of the cluster of galaxies Abell 383 with deep and a fairly homogeneous data coverage for the bands Z, J, H, and Ks. Figure 7.8 displays a Z+J+H pseudo-color composite of the study field. The two zoomed regions give an impression of the achievable image quality with OMEGA2000 under good observing conditions. The distortion-free camera optics in combination with the applied reduction schemes yield a homogeneously high image quality over the full FoV, which is demonstrated by the recovery of the subtle gravitational arc at the field edge in the lower right panel. The observing strategy of Sect. 7.2.2 aimed for a maximum net exposure time of 50 min per filter and field, in particular for the Z–H approach. The A 383 field was covered in all four bands to approximately this exposure time limit and is thus well suited for a cross- comparison. The data includes observations under photometric conditions for 58 minutes in Z, 50 minutes in J, and 50 minutes in the Ks-band. The 60 minute coverage of the H-band has been obtained under non-photometric conditions with partial cloud coverage, hence it provides only lower limits on the achievable depth. We follow Harris’ (1990) definition of a limiting magnitude as the magnitude at which the completeness of detected sources drops to 50% (50%-completeness limit). This limit can be robustly estimated from the differential number counts and corresponds approximately toa5 σ detection significance of the sources. The cumulative and differential number counts as derived for the central 192 square arcminutes (13.′8 13.′8) of the FoV are shown for all four filter bands in Fig. 7.9. From the differential number× counts in 0.1 magnitude bins in the left panels, the 50%-completeness limits are estimated as indicated by the dashed vertical lines. These limiting magnitudes are also shown for the cumulative number counts in the right panels. The normalization to square arcminute units directly provides the surface densities at a given magnitude limit

brightness and the number of individual images to be co-added. For both cases the expected optimization performance will increase with ﬁlter wavelength, i.e. in the order Z, J, H, Ks. 7.3. DEVELOPMENT OF A NIR SCIENCE REDUCTION PIPELINE 149

Figure 7.8: Z+J+H pseudo color image of the ﬁeld North of Abell383 used for the number counts in Fig. 7.9. Left: The full 192 square arcminute FoV considered for the analysis. The cD galaxy of Abell383 is just visible at the bottom center. Right: 4x4 arcmin zoom on the central part of the ﬁeld (top) and the Northern part of Abell 383 with its giant gravitational arc (bottom).

given by the dashed horizontal lines for the derived limiting magnitude in each band. The 50%-completeness limit for the Z-band in the top panels of Fig. 7.9 is determined as Z (AB) 23.9 corresponding to Z (Vega) 23.4. This limit confirms the expected lim ≃ lim ≃ high sensitivity of the NIR detector array in this last optical band. Comparing to Tab. 7.2 yields expectation values to what extent the Z-band galaxy luminosity function of a cluster will be probed at this maximum depth with respect to the characteristic magnitude m*. At z =1 this corresponds to m*+2, at z =1.5 to m*+0.1, and at z =2 to m* 0.8, revealing z 1.5 as the redshift where the maximum Z-band depth falls below the− characteristic ∼ magnitude. As a related consequence it can be expected that a possible red-sequence at z > 1.5 will be observable only with large scatter due to the increased photometric uncertainties∼ in the Z-band. The average background surface source density at the limiting Z-magnitude is 22 arcmin−2. At these densities even a bona fide cluster at intermediate redshifts as RX J0018.2+1617 in Fig. 7.6 reaches merely an overdensity of 40% above the background in the cluster center. For a 4′ 4′ region centered on the cluster, the overdensity drops to × zero and the cluster galaxies completely dissolve in the background of roughly 350 objects 150 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING within this sky area. This example illustrates the importance of (i) accurate positional information (5′′–10′′) as provided by the X-ray selection of candidates, and (ii) reliable color information in order to enhance the density contrast and allow a cluster identification. The number counts in the H-band are shown in the third row of Fig. 7.9. The 50%- completeness limit of H (Vega) 21.4 is routinely achieved in a good fraction of the lim ≃ XDCP fields even with shortened exposure times but better observing conditions concerning the transparency. This limiting magnitude is approximately 1 000 times fainter than the sky surface brightness. The H-band surface density of sources at this limit is 17 arcmin−2 and is thus about 20% lower than in the Z-band. At the same time the redshift grasp is greatly enhanced yielding a LF coverage to m*+2.3 at z =1 , m*+1.5 at z =1.5, and m*+0.9 at z = 2. This implies that the fraction of high-redshift objects is significantly increased in the H-band due to the very moderate redshift dimming of elliptical galaxies and the efficient relative suppression of blue foreground sources at these NIR wavelength. The second important conclusion is that the cluster luminosity function can be sufficiently probed out to redshifts of z 2. For the Z–H method this yields the performance expecta- ∼ tions that a sufficient number of cluster galaxies will be detectable in the H-band even at z > 1.5, but that the quality of the color information is decreased due to the photometric uncertainties∼ of the Z-band magnitudes. The situations in the J-band (second row) and Ks-band (bottom row) are similar to H and can both probe the cluster luminosity function to about m*+1 at z = 2. The deep J-band limit of J (Vega) 23.1 exhibits a surface density of >30 arcmin−2. The Ks lim ≃ observations suffered from dome-scattered moonlight in a good fraction∼ of the FoV which manifests itself in the suspicious upturn18 of the number counts at Ks>20. The counts are thus only trustworthy up to a number density of 20 arcmin−2 at Ks∼ 20. A realistic ∼ ∼ corrected limiting Ks magnitude is approximately Kslim(Vega) 20.5. In any case, the J–Ks method is capable of providing accurate color information out≃ to the highest redshifts and is thus very valuable in conjunction with the Z–H strategy for the identification of the most distant clusters to be found. In summary, we have shown that the achievable data quality and reduction performance fulfills the depth requirements and expectations of Sects. 7.1.2 & 7.2.2. The applied imaging strategy with maximal net exposure times of 50 min per filter and field is sufficiently deep ∼ to provide stand-alone Z–H colors out to z 1.5 and has the potential to go significantly beyond this limit with complementary J and∼ Ks observations. The expected surface density of background sources of typically 20 arcmin−2 in deep XDCP fields emphasizes the need ∼ for accurate color information. The procedures for obtaining the final photometry, object catalogs, and pseudo color images from the reduced data will be introduced in the next section.

18The highest positive spatial ﬂuctuations of the additional scattered light surpass the detection threshold and are falsely interpreted as faint objects. 7.3. DEVELOPMENT OF A NIR SCIENCE REDUCTION PIPELINE 151

Figure 7.9: Z, J, H, and Ks number counts of the field displayed in Fig.7.8. Z (58 min), J (50 min), and Ks (50 min) were observed in photometric 1′′-seeing conditions, the H-band (60 min) observations were not photometric and suffered from partial cloud coverage, i.e. the limit is a lower bound. The left panels show the differential number counts in 0.1 magnitude bins. The 50%-completeness limit is given by the dashed vertical line. The log N–log S distributions are displayed in the right panels, with the horizontal dashed lines indicating the average total object density at the limiting magnitude. 152 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

NIR Photometric Analysis

H Z Z Z Z sum sum photom. photom. photom. single single single image image image 1 image 2 image 3

Astrometry

Image Alignment in pixel coordinates

Variance Weighted Detection Image

Seeing Adjustment 2MASS point Astrometric Calibration source World Coordinate System catalog

Photometry H-band H-band photometric Sextractor Photometry catalog

calibrated Z Zeropoint Calibrated Z-Band from standard star Z-band observations Reference Star Catalog reference objects

Z-Band Deep Field Z-band Sextractor Photometry photometric catalog

Analysis

Pseudo Color Images

Color-Magnitude-Diagrams

Combined X-ray-Optical Optical Richness & X-ray Morphology Source Evaluation Estimates

Model Comparison & Redshift Estimate Redshift Estimate & Lx, Mass Estimates

Figure 7.10: Near-infrared photometric analysis ﬂow chart. 7.4. PHOTOMETRIC ANALYSIS 153

7.4 Photometric Analysis

With the ﬁnal optimized data products from the NIR reduction process in hand, we can now turn to the photometric analysis. The general objective will be the measurement of the cluster red-sequence location with an absolute color accuracy of σcolor 0.1 mag in order to achieve the redshift accuracies (σ 0.1) as discussed in Fig. 7.2. ∼ z ∼

7.4.1 General procedure The generic photometric analysis procedure as implemented for the XDCP follow-up program is illustrated in the flow chart of Fig. 7.10. The analysis process from the image level to final redshift estimates can be subdivided into three main modules: (i) The astrometry part during which images are aligned, combined, seeing adjusted, and finally transformed to sky coordinates, (ii) the photometry module dealing with object detection, characterization, and calibration for all filter bands, and finally (iii) the analysis and evaluation section based on composite color images, color-magnitude diagrams, and comparison to model predictions. The discussion will focus on the constituents of the Z–H method, however, the procedures for complementary J or Ks-band data are analogous to the H-band analysis.

Astrometry Accurate image alignment in pixel coordinates is a pre-requisite for the source extraction software SExtractor (Bertin and Arnouts, 1996) and is performed as a first step. Input images for the alignment procedure are the final deep summed images in the H-band and Z- band, plus additionally three single Z-band images acquired under photometric conditions for the calibration process. In practice, the H image is used as master grid, and all Z- band images are integer-pixel19 aligned in sky coordinates via object matching and saved with the new coordinate grid. Since this procedure is similar to the summation process in Sect. 7.3.2, the NIR pipeline can be tuned to provide the accurate automatic object alignment in conjunction with a meta script organizing the proper image handling. During the same process a combined deep detection image DETZH is created from the variance-weighted sum of the Z and H-band frames as DET = Z σ−2 + H σ−2, ZH · Z · H where σZ and σH are the measured global background standard deviations in the frames. This detection frame serves the three main objectives (i) to maximize the depth and the number of detectable objects, (ii) to provide averaged object profiles for an unbiased20 magnitude and color determination, and (iii) to be used in the green (neutral) channel for color-composite images. This detection frame will be handed over to SExtractor for finding sources and determining their source apertures, while the actual measurement is performed in the individual H or Z image. The variance-weighted image combination makes direct

19Fractional pixel oﬀsets cannot be applied in this case since the Z-band images are only aligned to the master grid, but not co-added. 20Due to possible positional truncation errors during the alignment, detecting sources in H and measuring the magnitude in Z bears the risk of a systematic bias towards fainter Z magnitudes. 154 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

use of the deep frames as provided by the NIR pipeline in order to achieve a maximized detection depth, even though the frames might still exhibit different seeing levels. Since there will be no color information without an initial detection of a source, the source detection depth is given priority over the color accuracy. A fixed measurement aperture applied to images observed under discrepant seeing conditions probes different physical scales of the same galaxy, e.g. a 1.5′′ aperture might contain most of the (distant) galaxy light under good conditions while under bad seeing the measured flux will be dominated by the galaxy center. Since elliptical galaxies often exhibit color gradients as a function of distance from the center, the same physical scales need to be considered for an unbiased and well-defined color measurement. The potentially discrepant seeing conditions are adjusted in the next step, where additional smoothing with a Gaussian kernel of radius rsmooth is applied to the image with better seeing σgood. Since the standard deviation of the Gaussian kernels add in quadrature, the smoothed frame will 21 have the same effective seeing σbad, measured as the full-width-at-half-maximum , if the condition σ2 = σ2 +(2.36 r )2 for the smoothing radius r is fulfilled. Mea- bad good · smooth smooth suring the color in these seeing adjusted images ensures a flux extraction from the same physical galaxy scale while maintaining the maximized depth of the detection image. As a final preparatory step, the aligned and seeing-adjusted images are equipped with a proper equatorial world coordinate system (WCS). The transformation equations from XY image coordinates to the right ascension (α or RA) and declination (δ or DEC) system are obtained by finding the astrometric plate solution from sources in the FoV with accurately known absolute positions. In practice this task can be performed with the WCS Tools22 software package which automatically queries the 2MASS point source catalog for reference sources, finds and matches objects, solves for the plate solution via least-square fitting, and updates the image header with the appropriate WCS keywords. Details on the exact transformations and the representation of the world coordinate system in the FITS image header can be found in Calabretta & Greissen (2002).

Source detection and characterization We can now turn to the actual objective of the imaging enterprise, the detection and photometry of astronomical sources with the focus on faint distant galaxies. This central task is performed with the help of the widely used source extraction and characterization software SExtractor. This versatile photometry code detects, de-blends, measures and classifies sources from astronomical images and saves the results in a photometric catalog for further processing. For the science objectives of deep imaging and accurate colors we make use of SExtrac- tor’s dual image mode. In this configuration, the object search and structural characterization takes place in the depth-optimized variance-weighted detection image, from which the final source list and the individual object extraction regions are determined. The ac-

21The relationship between the standard deviation σ and the full-width-at-half-maximum of a Gaussian proﬁle is FWHM= 2.36 σ. 22The free software is· available at http://tdc-www.harvard.edu/software/wcstools. 7.4. PHOTOMETRIC ANALYSIS 155

tual photometry measurements, i.e. the final source flux determinations, are subsequently performed in the aligned individual filter band images, Z or H, using the exact object position and extraction region information of the input source list. This technique ensures that all objects in all filters have the same photometry areas and apertures appropriate for measuring colors and eases the cross-identification of objects in different catalogs. The pre-requisites for this method, aligned images with the same size, have been fulfilled in the astrometry module. The last requirement of coincident sky areas of the different input images is met by providing external variance maps to SExtractor which exclude the non-overlapping dither and pointing offsets of the images from the photometric analysis. For the overlapping region of interest, the variance maps contain estimates of the local noise levels for the individual images. These noise maps are obtained from an empirical variance measurement in a central statistics window divided by the smoothed flatfield to account for variations due to large-scale changes of the detector quantum efficiency. The detection and measurement images, variance maps, the photometric zero point, seeing and saturation level estimates, and other data characteristics are then pipelined to SExtractor for the photometry. In short, SExtractor performs the following steps: (i) estimate the background and its local noise properties, (ii) determine whether pixels belong to background or objects, (iii) split up object areas into separate objects, and (iv) determine the properties of each object. A critical component for the detection performance is again the quality of the background estimation, which should be determined from a fine enough grid to account for large-scale variations and at the same time robust enough to avoid local biasing from nearby bright objects. The second crucial set of parameters is related to the threshold for an object detection. As minimal object significance, the implemented threshold requires at least four connected pixels with values higher than one standard deviation above the background noise, which implies a 2-σ detection for the limiting case. Since the main scientific interest focusses on the spatial and color distribution of cluster galaxy populations rather than individual faint galaxies, the expected spurious source rate on the percent level is an acceptable trade-off for a deeper coverage. The second shortcoming of the implemented approach is a less accurate determination of the structural object parameters23 from the combined, and thus averaged, detection image, which is again outweighed by the optimized depth and color information. Finally, the photometric precision and error estimates are compromised somewhat by performing the measurements in the deep sum images, instead of the individual single images, which would allow the object properties to be determined from an ensemble of measurements. The latter approach enables e.g. the elimination of cosmic ray events inside the object apertures, but the expected marginal gain does not justify the significant extra effort for red-sequence-based methods. The last consideration concerning SExtractor photometry deals with relating the measurements to the galaxy model predictions. The models specify total apparent object magnitudes for the different filters, which are best matched by the MAG AUTO magnitude, for which SExtractor applies automatic flexible elliptical apertures to each object in order

23 The expected median image alignment oﬀset of about 0.1′′ between the H and Z frames introduces a slight object ‘smearing’, which decreases the quality of the stellarity index measurement, i.e. the point source identiﬁcation. 156 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

to estimate the total ﬂux. For an accurate object color determination, however, the use of a smaller aperture containing only pixels with a suﬃciently high signal-to-noise ratio is advantageous over measurements including the low SNR object periphery. This can be achieved in practice through MAG ISO magnitudes for the Z–H color determination. These isophotal magnitudes are measured from all connected object pixels above the detection threshold.

Analysis and redshift estimation The analysis and candidate evaluation represents the last module of the NIR imaging procedure. For this all available information on the cluster candidate are combined into final data products. For the visual cluster identification, a pseudo color composite image is created from all filter bands. In the common case with only Z and H-band data available, the detection image is used for the third color channel. Color images in general are obtained by overlaying a red, a green, and a blue data channel (RGB colors). Applied to our case, the Z-data is loaded into the blue channel, the detection image into the (neutral) green channel, and the H-band is represented in red. The combined imaging data can be complemented by overlaying the matching X-ray contours and the extracted photometric information for each object. As the second important diagnostic tool, color-magnitude diagrams are created from the photometric Z and H-band catalogs for the quantitative analysis. The pre-matched object catalogs are read, the Z-magnitudes transformed to the Vega system, and the Z–H isophotal object colors determined. These are then plotted against the total H-magnitude for all objects, and sources within 30′′ and 1′-apertures around the centers of cluster candidates are highlighted. As an additional diagnostic the object overdensity with respect to the average background density is computed for the given apertures. No special efforts are taken to separate galaxies from stellar sources since (i) stars are typically easily rec- ognizable in the color composites and can be appropriately taken into account if located close to a cluster, (ii) SExtractor’s stellarity classification is less reliable when applied to the variance-weighted detection image, and (iii) the automatic classification breaks down at faint magnitudes anyway. If a red-sequence is identified for a cluster candidate, the best-fit color estimate can be directly transformed into a redshift estimate based on a formation redshift zf =5 model with solar metallicity (see Sect. 10.1) as shown in the lower right panel of Fig. 7.1. Since the follow-up photometric accuracy is typically not sufficient to trace the red-sequence slope for high-redshift clusters, one has to aim at an average red-sequence color which will be naturally hinged at m* galaxies as a result of the achievable depth limit. Combining the cluster redshift estimate with the X-ray flux measurement, an approximate X-ray luminosity LX and a preliminary mass proxy can be obtained based on the LX –M scaling relation. The relative galaxy overdensity and red-sequence population additionally yield a first clue on the optical richness of the cluster, and the features in the X-ray morphology might hint at the dynamical state of the system. For an extended cluster 7.4. PHOTOMETRIC ANALYSIS 157

candidate evaluation, the following cluster features can be considered: (i) the presence of a central dominant galaxy (see Sect. 11.1), (ii) a systematic BCG offset from the X-ray center (see Fig. 11.1), (iii) a possible AGN contribution to the X-ray emission, or (iv) the potential coincidence of galaxy and X-ray features such as filamentary extensions (see Sect. 10.2). We have finally arrived at the initial objective of the two-band follow-up imaging strategy. As survey step 3, the presence of a cluster can be confirmed based on the observed overdensity of red galaxies, and for step 4 a redshift estimate is obtained from the color of the red-sequence. Promising high-redshift candidates are now ready for spectroscopic confirmation as discussed in Chap. 8.

7.4.2 Photometric calibration We will now come back and discuss the photometric calibration procedure, which was left out in the last section. Due to the different nature of the optical Z-band and the NIR H-filter, the calibration methods differ in both band-passes and will be introduced separately.

2MASS NIR calibration The 2-Micron All-Sky Survey (2MASS) (Cutri et al., 2003) provides full sky coverage in 24 the J, H, and Ks-band to a minimum depth of Jmin(Vega)=15.9, Hmin(Vega)=15.0, and Ksmin(Vega)=14.3 for 10-σ point source detections. The availability of this high quality data set allows a direct and easy cross-calibration to the 2MASS photometric systems via serendipitous point sources of sufficient brightness in OMEGA2000’s large FoV. Since the 2MASS calibration stars with known calibrated magnitudes are in the science data and are only extracted after the observations, they have experienced the same atmospheric extinction as all sources in the field. This implies that airmass corrections as for the Z-band are not needed, and that a calibration is even possible under non-photometric observing conditions. Homogenous extinction across the 15′-FoV is the only assumption that is made for the case of low or varying transparency, e.g. due to moderate cloud coverage. This assumption is justified in most cases and can be cross-checked by testing for systematic photometric offsets across the FoV. Hence, the H-band calibration has low requirements on the conditions and allows continuing observations even under challenging circumstances. In practice, the calibration process is achieved through the following steps. (i) The raw photometric H-band catalog25 of the science field is matched to the 2MASS catalog of the bright (H<14.5 mag) point sources in the FoV as was obtained during the astrometric calibration. (ii) The zero point offset of each source is determined by subtracting the instrumental 4′′-aperture magnitudes from the calibrated H-band object magnitudes in the

24See http://www.ipac.caltech.edu/2mass/releases/allsky/doc/sec2 2b.html. Minimum depth means that this value is achieved for the full survey area, while parts of the sky have a deeper coverage. 252MASS data products can be downloaded from http://irsa.ipac.caltech.edu/applications/ Gator/. 158 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

reference catalog. (iii) The average effective zero point (ZP) for the observation and its standard deviation is computed via iterated outlier clipping of all calibration sources. (iv) This ZP is provided to SExtractor as an additive offset for the second processing which will yield calibrated magnitudes. The calibration process for a sample field is illustrated in the upper right panel of Fig. 7.11 which shows the distribution of the individual calibration source zero points. The typically 20–80 serendipitious bright 2MASS stars per field provide a zero point accuracy of approximately σ 0.02–0.04 mag. The upper right panel of Fig. 7.11 demonstrates ZP ≃ that the determined zero points do not exhibit a systematic trend with the object color, i.e. a color term correction is not necessary.

Z-band calibration Unfortunately, a 2MASS-like all-sky survey in the Z-band does not exist yet resulting in a photometric calibration process for this filter which requires more effort. The Sloan Digital Sky Survey (SDSS), with 10 000 square degree multi-band imaging coverage in the Northern hemisphere, overlaps with about 15–20% of the XDCP survey fields and thus only allows calibration cross-checks but no substitute for designated standard star observations. In order to achieve an accurate calibration to the SDSS photometric system, several Z-band standards from the Smith et al. catalog (2002) were observed at different airmasses over the course of each photometric night. Since a good fraction of the Z-band science observations were performed in non-photometric conditions, we obtained additional short image series for these fields during the photometric nights resulting in a minimum of five available frames per field for the calibration process. This photometric overlap subsequently allows ‘bootstrapping’ of the calibration for the deep science observation as will be discussed. Since no experience with the OMEGA2000 Z-band calibration existed before the XDCP program, all necessary calibration information had to be derived from the data. The following photometric transformation equation is considered

Z = Z + A A X + A (R Z) , (7.5) cal instr 0 − 1 · 2 · − 26 where A0 is the nightly photometric zero point , A1 is the Z-band atmospheric extinction coefficient per unit airmass, and A2 is a potential color term that could be introduced by a different filter-detector response relative to the SDSS system. The extinction term A X − 1· corrects the measured magnitudes to the original flux values above the atmosphere, i.e. the lost flux due to atmospheric extinction is added.

26 1 The instrumental magnitude is given by minstr 2.5 log(counts s− ). Hence, the zero point is in eﬀect the apparent magnitude of a source for which≡− the instrument-telescope system detected one count per second. 7.4. PHOTOMETRIC ANALYSIS 159

Figure 7.11: Photometric calibration. Top row: Photometric zero point calibration in the H-band using bright 2MASS reference point sources in the field-of-view (left). The large FoV usually contains several dozen suitable calibration stars for any pointing and allows a zero point calibration to typically σ 0.03 mag. The increased scatter towards the faint end reflects the photometric uncertainties of H ∼ the shallower 2MASS survey and is the reason for only selecting calibration reference sources with H<14.5. The right panel shows the zero point offset from the mean as a function of the object H–Ks color and illustrates that a color term correction with respect to the reference 2MASS system is negligible. Bottom row: Z-band calibration to the SDSS photometric system. Since most XDCP fields do not have direct SDSS coverage, the nightly zero points have to be obtained from Z-band standard star observations, with which a calibrated reference star catalog can be obtained from short photometric exposures of the science field. The catalog is then used to calibrate the zero point in the deep Z-band field in the same manner as for the H-band (left panel). The color term check (right panel) also yields a correction term that is close to zero. 160 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING

The available standard star observations were not sufficient in quality and quantity to solve the transformation equation simultaneously for the three unknown parameters A0, A1, and A2. We thus choose the following approach: (i) Determine the color term coefficient A2 from the best suitable field with SDSS27 coverage. (ii) Measure the extinction coefficient A1 with fixed color term A2 for the photometric night with best standard star coverage and use this value for all nights. (iii) Compute the photometric zero point A0 for each photometric night using the pre-determined values for A1 and A2. The bottom right panel of Fig. 7.11 displays the zero point offsets as a function of (R– Z)-color for a SDSS calibration field with about 100 reference stars. The color coefficient A is obtained from the slope of the linear regression analysis yielding A = 0.014 0.01 2 2 ± and is thus consistent with zero within the errors. This formal result is good news confirming a posteriori that the Z-band NIR instrument response is very similar to the optical SDSS system. The small color term correction on the percent level thus eases the direct comparison of the OMEGA2000 Z-band magnitudes to SDSS reference values. With the color term fixed at the above value, the extinction coefficient A1 is determined from data of the highest quality night (10 November 2006). The instrumental magnitudes minstr for the observed standard stars are plotted against the airmass under which the data was taken. The average slopes of these relations for each standard star then yield the Z-band atmospheric extinction coefficient A1 =0.045 0.01 per unit airmass. This result is consistent with the values obtained at other observatori± es and implies that the instrumental Z-band magnitudes are only dimmed by 0.05 mag for observations at airmass two ∼ compared to zenith pointings. This moderate extinction, which continues to decrease in the NIR bands, is a manifestation of the atmospheric Rayleigh scattering with cross-section σ λ−4. The small extinction correction of A X 0.05–0.1 also justifies keeping the R ∝ 1 · ∼ coefficient A1 at a fixed value for all nights. The last unknown parameter for the photometric transformation of Equ. 7.5 is the nightly zero point A0. The zero point is obtained by applying Equ. 7.5 to all standard star observations of a given night and using the above values for A1 and A2. Out of the seven photometric nights used for the Z-band calibration, estimated zero point errors of σZP 0.02–0.03 mags were achieved for five nights; two nights of service observations turned out∼ to be non-photometric with estimated zero point uncertainties of σ 0.2 mags. ZP ∼ We now have all the tools at hand to perform the photometric ‘bootstrap’ calibration of the deep Z-band science fields as illustrated in Fig. 7.10. The approach uses the following steps: (i) The astrometric preparatory module is applied to three single photometric snap- shot observations in the same way as the deep Z-band science image, i.e. the images are aligned and seeing adjusted. (ii) SExtractor source detection is performed on these calibration images in dual image mode with the deep detection image as reference. (iii) The magnitudes28 of the three resulting object catalogs are photometrically calibrated using Equ. 7.5 with the appropriate nightly zero point. (iv) Bright but unsaturated stellar

27SDSS data release 5 (DR5) archive data products can be retrieved from http://cas.sdss.or. /astodr5/en/tools/search/IQS.asp. 28 The SDSS model magnitudes for mcal of the calibration stars are matched to the measured total MAGAUTO magnitudes for minstr. 7.5. SUMMARY 161

sources with 14.5

After this long and comprehensive chapter on near-infrared imaging, the main achievements within this thesis work are now summarized:

Development of a new NIR-based Z–H follow-up imaging strategy; • PI of a two-semester 10-night imaging proposal to Calar Alto and award of both • observing runs with class A priority;

Led observations at the Calar Alto 3.5 m telescope with 24 nights of on-site presence; • Co-led a three-night NIR imaging run at the CTIO 4 m Blanco telescope which com- • plemented an optical follow-up program with ten nights of on-site presence (results not included in thesis);

Development of a versatile NIR science-grade reduction pipeline; • Development of science analysis tools for red-sequence applications; • Reduction and analysis of the data from all NIR imaging runs. • 162 CHAPTER 7. NEAR-INFRARED FOLLOW-UP IMAGING Chapter 8

Spectroscopic Follow-up

The spectroscopic confirmation of distant cluster candidates is step 5 of the survey strategy discussed in Sect. 5.2 and constitutes the observational bottleneck for cluster studies at high redshift. At z > 1, spectroscopy of cluster galaxies is currently only feasible with very few instruments at∼ 8–10 m-class telescopes, and at z > 1.5 any study is extremely challenging with the available technology. ∼ The final objectives of the spectroscopic analysis in survey mode are: (i) the confirmation of the cluster as a three-dimensionally bound object, (ii) a precise determination of its redshift, (iii) identification of cluster member galaxies, and (iv) a first estimate of the cluster velocity dispersion. A combination with the available X-ray properties of the cluster enables the derivation of the X-ray luminosity and thus a mass proxy (see Sect. 2.3.4). These physical quantities are crucial for linking the observations to theoretical predictions and hence for the use of clusters as cosmological probes. Further in-depth investigations for individual selected systems (step 6) allow (v) the study of the cluster dynamics, (vi) a substructure and LSS analysis, and (vii) detailed investigations of the cluster galaxy population such as derivations of the star formation history, age, and metallicty of individual galaxies. Since the spectroscopic work for this thesis has been less extensive than the observational data described in the previous sections, this chapter is intended to provide a compact overview1 of the basic spectroscopic methods and applications.

8.1 Basics of Spectroscopy

Simply speaking, an optical spectrograph consists of (i) a slit aligned to the target in the focal plane, (ii) a collimator, (iii) a light dispersing element, (iv) a camera lens, and (v) a detector. Many modern spectrographs use grisms as dispersing elements, which consist of a combination of a transmission grating and a prism. The latter counteracts the redirection from the optical axis and thus allows compact and linear on-axis arrangements. The

1A more detailed description of the reduction process can be found in Novak (2003).

163 164 CHAPTER 8. SPECTROSCOPIC FOLLOW-UP principal diﬀraction maxima of order m at angle θ for light rays with an incident angle i obey the grating equation

m λ = d [sin(i) + sin(θ)] , (8.1) · max · where d is the grating constant (slit distance). Taking the focal length ratio fcam of the optical system into account, this translates into the linear dispersion ∆l in the detector plane

∆l m = f . (8.2) ∆λ cam · d cos(θ) · The resulting spectral resolving power R of the grating is then given by the expression

λ R = = m N, (8.3) ∆λ · with N being the total number of lines. Depending on the science application, the grisms can be optimized (blazed) to yield the maximum throughput at a given central wavelength λcen for the desired spectral resolution R. The XDCP science goals aim at multi-object spectroscopy of very faint, high-redshift galaxies which requires the use of red-sensitive spectrographs at 8 m-class telescopes. The XDCP instrument-of-choice is the FOcal Reducer and low dispersion Spectrograph for the Very Large Telescope, or short VLT FORS 2 (Jehin, 2006), a first generation VLT multi-purpose instrument which has been refurbished with red-optimized CCDs providing an average quantum efficiency of about 65% in the critical Z-band. FORS 2 offers multi- object spectroscopic capabilities with exchangeable laser-cut slit masks (MXU unit) for a simultaneous coverage of about 30 target spectra within the 6.8′ 6.8′ field-of-view. The × GRIS 300I+21 grism with a central wavelength of λcen =8600 A˚ and a peak efficiency of 70% achieves a wavelength coverage of 6 000–11 000 A˚ with a resolution of R=660 or 13 A˚ (corresponding to four binned pixels). This setup optimizes the signal-to-noise ratio of the target continuum flux, while still resolving the major spectral object features and the sky emission lines. The most prominent optical spectral features of galaxies are listed in the left panel of Tab. 8.1. For galaxies without active nuclei, the spectra reflect the underlying dominant stellar populations which depend on the star formation history. As discussed in Sect. 2.4, the centers of clusters are characterized by the highest density of early-type galaxies, with mostly ‘old, dead, and red’ stellar populations dominated by red giants. Owing to the absence of recent star-formation, elliptical and lenticular galaxies exhibit predominantly absorption features in their spectra. The most prominent spectral characteristic of early- type galaxies is the D4000 break around 4000 A.˚ This break is caused by a complex interplay of a large number of absorption lines (e.g. Fe I, Mg I, and Balmer lines) in conjunction with the declining continuum of cold stars towards shorter wavelengths. Most notable are the strong H & K resonance lines of single-ionized Calcium originating in the stellar atmospheres of G and K-giants. The D4000 break requires only a moderate signal-to-noise 8.2. CHALLENGES OF DISTANT GALAXY SPECTROSCOPY 165

Feature Rest Wavel. [A]˚ Type E/S Sky Feature Wavel. [A]˚ Type

Mg II 2800 a E, S O2 B-band 6867–6919 a [O II] 3 727.0 e S O2 A-band 7607–7633 a Ca II K 3 933.6 a E, S H2O 9400–99700 a Ca II H 3 968.5 a E, S [O I] 5 577.3 e Hδ 4 101.7 e or a S Na I D 5 892.5 e CN G-band 4 300.4 a E [O I] 6 300.3 e Hγ 4 340.5 e or a S [O I] 6 363.8 e Hβ 4 861.3 e or a S OH− 6820–7000 e [O III] 4958.9, 5006.8 e S OH− 7240–7580 e Mg I 5 175.4 (blend) a E OH− 7720–8100 e Na I D 5 892.5 (blend) a E OH− 8290–8500 e Hα 6 562.8 e S OH− 8760–9100 e [S II] 6717.0, 6731.3 e S OH− 9300–10200 e

Table 8.1: Optical spectral galaxy and night sky features. Left: Rest wavelength for the most prominent optical absorption and emission lines for early and late-type galaxies. The line type indicates emission lines (e) or absorption lines (a), E (S) means that the spectral feature is predominantly found in elliptical (spiral) galaxies. Right: Important telluric absorption lines (top) and sky emission lines (bottom) as shown in Fig. 8.1. Brackets around the element name mark forbidden transition lines. For more details see Cox (2000).

ratio for identification, shows little contamination from reddening, and does not require absolute fluxes, hence it is the most important redshift indicator for early-type galaxies. Spiral galaxies have experienced a more recent star formation epoch and thus have hotter stars contributing to the spectra. Their D4000 breaks are less pronounced than in ellipticals and can exhibit emission lines (e.g. Hα, Hβ, [O II], [O III]). Also common in clusters are galaxies with active nuclei, e.g. Seyfert-type galaxies, which show very strong emission lines. Sample spectra of typical low-redshift cluster galaxies with identified features have been shown in Fig. 4.5.

8.2 Challenges of Distant Galaxy Spectroscopy

The spectral galaxy features with rest-frame wavelength λ0 are redshifted by the cosmological expansion to the observed wavelength λobs given by

λ = (1+ z) λ . (8.4) obs · 0 For z > 1 objects this implies that the important D4000 break of early-type galaxies is shifted∼ into the I and Z-band, limiting the number of spectral features usable for a secure redshift determination to a few lines with λ0 <4500 A˚ (see Tab. 8.1). At z >1.5 the D4000 ∼ 166 CHAPTER 8. SPECTROSCOPIC FOLLOW-UP break is completely lost for optical observations, which marks the onset of the so-called ‘redshift-desert’. Besides the decreasing number of accessible spectral features, the second observational challenge is posed by the onset of atmospheric contamination in the I-band and in particular in the Z-band. Figure 8.1 illustrates the telluric2 absorption (left panel) and the atmospheric emission (right panel) in the wavelength range from 6000 to 10000 A.˚ The visible telluric absorption features in the left panel are caused by O2 and H2O. The strong oxygen A-band around 7 620 A˚ and the B-band at 6900 A˚ can easily mimic galaxy absorption features and have to be carefully excised, in particular if object lines are nearby. Towards the NIR, water vapor becomes the principal absorber with the first strong absorption band starting at around 9400 A.˚ The atmospheric emission in the right panel depicts the spectrally resolved reason for the background challenge in the near-infrared, the hydroxyl OH− emission. These radicals are produced during the day in an excitation reaction of ozone O3 and hydrogen H at an altitude of 85–100 km and radiate away their excess energy during the night. The airglow emission becomes dominant beyond about 7 000 A˚ and is the limiting factor for ground based observation throughout the NIR regime. The so-called Meinel bands are easily discernable and correspond to different vibrational molecular transitions, whereas the lines within a band originate from rotational dipole transitions. For comparison, the strong emission lines beyond 8 000 A˚ have peak fluxes roughly 1 000 times brighter than the continuum of typical z 1 galaxies. Sky line residuals will thus manifest a noise limit ∼ for the identification of galaxy spectral features. In order to obtain a sufficiently high signal-to-noise ratio for passive cluster galaxies of candidates at redshifts 1 < z < 1.4, typical exposure times of 2–4 hours are required even with VLT FORS 2. XMMU∼ J2235-2557∼ sample spectra at z 1.4 were shown in Fig. 5.1; ∼ additional z 1 galaxy spectra will be discussed (see Fig. 10.5). ∼ Efficient strategies for the spectroscopic confirmation of the highest redshift cluster candidates at z > 1.5 have yet to emerge and are currently at the limit of the technological capabilities.∼ In principle, there are two approaches to tackle the problem of z> 1.5 redshifts, (i) the classical optical spectroscopy with extremely long exposure times to enable the identification of the weak rest-frame UV features, or (ii) moving to near-infrared spectroscopy in analogy to the imaging strategy. The optical multi-object spectroscopy approach has been used to probe the general galaxy population in the ‘redshift-desert’ for the Gemini Deep Deep Survey (GDDS) (Abra- ham et al., 2004) targeting galaxies with redshifts 1–2. Even with typical exposure times of 20–30 hours per field as for the GDDS, a secure redshift determination for passive galaxies is very hard once the D4000 is shifted beyond the probed wavelength range. Owing to the extreme exposure time requirements, the optical spectroscopy approach is not well-suited for routine cluster confirmations of the highest redshift candidates, but is invaluable up to z 1.5. ∼

2Telluric means in Earth’s atmosphere. 8.2. CHALLENGES OF DISTANT GALAXY SPECTROSCOPY 167

Figure 8.1: Spectral absorption and emission features introduced by the night sky. Left: Normalized telluric absorption lines as determined from the analysis of an intrinsic featureless spectrum of a white dwarf. The strong absorption bands around 6 900 A˚ and 7 620 A˚ are caused by atmospheric O2. Right: Night sky emission spectrum dominated by the OH− radical at wavelength above 6 800 A.˚ The diﬀerent molecular Meinel-bands are easily visible. The apparent weakening towards longer wavelength is caused by the decreasing response of the camera system; the intrinsic line brightness continues to grow drastically towards the NIR regime. The lines above 8 000 A˚ exhibit peak ﬂux levels which are approximately a factor of 1000 above the continuum level of typical distant cluster galaxies.

Near-infrared spectroscopy could in principle take over once the 4000 A˚ rest-frame features are redshifted into the NIR, i.e. at around z 1.4–1.5. However, several observational and technological issues turn this approach also into∼ a challenging enterprise. (i) The wavelength regimes outside the J, H, and K-band are heavily absorbed by the atmosphere, i.e. spectra are not homogeneously probed in wavelength. (ii) The sky emission lines within the NIR bands are so strong and numerous that the identification of the galaxy continuum and its absorption features is very difficult. (iii) Current NIR spectroscopy instruments typically exhibit significant trade-offs between wavelength coverage, spectral resolution, and multi-object capabilities, i.e. for optimal instrument settings single object spectroscopy might be required. The technological aspect of the last item will improve significantly over the next few years with several powerful NIR multi-object spectrographs under development or in the commissioning phase. Their performance and application concerning high-redshift passive galaxies with spectral absorption features only is yet to be tested carefully. A possible modified strategy in the NIR regime would be the emphasis on bluer emission line cluster galaxies, resulting in an easier redshift determination at the expense of a less secure object pre-selection, i.e. a significant fraction of blue objects will be foreground sources. In summary, the spectroscopic confirmation of galaxy clusters at z> 1 will always be the observational bottleneck for XDCP-like projects. The best currently available optical spectrographs allow routine identifications in survey mode, i.e. with reasonable exposure 168 CHAPTER 8. SPECTROSCOPIC FOLLOW-UP times, out to z 1.4–1.5. Moving beyond this limit into the ‘redshift desert’ is extremely challenging and≃ will likely require the best new NIR multi-object spectrographs of the near future. In any case, the spectroscopic demands posed by the highest redshift clusters are at the technological limit of 8–10 m-class telescope and will remain challenging for at least another decade.

8.3 Spectroscopic Data Reduction

In contrast to the reduction software developments in Chaps.6&7, the optical spectroscopic data reduction introduced in this section is standard and did not require an extensive customization beyond the available IRAF 3 reduction routines. Therefore the discussion is kept short with the main aim of providing an overview of the different steps from raw data to the final spectroscopic redshifts. Figure 8.2 illustrates the principal reduction steps of optical spectroscopy in a flow chart with IRAF packages and the main routines indicated. The reduction procedure can be subdivided into three main modules: (i) the general CCD image reduction, (ii) the extraction and calibration of the 1-D spectra, and (iii) the redshift analysis. The CCD image reduction, as the first module, is similar to some of the basic steps discussed for the NIR pipeline in Sect. 7.3. As additional preparatory steps for CCD images, the frames have to be first trimmed to extract the actual data section, and then bias subtracted in order to remove the pedestal count level featured by CCDs. The additive applied bias voltage offset can be measured from the overscan CCD regions and designated bias frames taken with zero exposure time. The reconstructed master bias frame is then subtracted from the data to yield intermediate products which contain the measured signal values. As a second calibration step, the multiplicative flatfield correction is applied by dividing through the fixed-pattern-noise frame. In contrast to the NIR imaging flatfield, which included the global quantum efficiency variations, the spectroscopic flatfield only contains the local pixel-to-pixel variations since the full spectrum is now spectrally dispersed over the CCD length. In practice, the local pixel-to-pixel sensitivity variations are extracted by dividing the global spectroscopic dome flatfields, taken with the proper slit or mask configuration, by a smooth continuum fit along the dispersion axis to remove the global large-scale modulation4 and keep the local median levels to unity. In case a proper flux calibration is required (which is often not necessary for a redshift determination) a correction function along the dispersion axis can be constructed from observations of a spectroscopic flux standard star with accurately known spectral properties. The CCD image reduction is completed by applying the bad pixel correction and summing all individual frames per science target (typically three) to a cosmics cleaned final sum image, in analogy to the NIR imaging procedure.

3For software and documentation see http://iraf.noao.edu. 4The large-scale variations are caused by the detector QE function and the unknown spectral shape of the ﬂatﬁeld lamp. 8.3. SPECTROSCOPIC DATA REDUCTION 169

Optical Spectroscopy with IRAF

bad raw λ-CAL bias master images frame ﬂatﬁeld pixel images mask

CCD Image Reduction ccdred

Bias Subtraction ccdproc Flatﬁeld Correction ccdproc Bad Pixel Correction ccdproc

Image Summation & Cosmics Removal imcombine

Calibrated 1-D Spectra specred

1-D Spectrum Extraction & Background Subtraction apall

Line Identifcation & Dispersion Solution identify

Wavelength Calibration dispcor

Cosmetic Corrections calibr. splot spectra

Redshift Determination astutil & rvsao

Emission Line Removal lineclean

Heliocentric Vel. Correction rvcorrect

Cross-Correlation Analysis xcsao galaxy redshifts

Figure 8.2: Principal data reduction steps for optical spectroscopy with IRAF. The employed IRAF packages and main tasks are indicated in the corresponding module boxes. 170 CHAPTER 8. SPECTROSCOPIC FOLLOW-UP

The second module of the reduction procedure in Fig. 8.2 has the main objective of obtaining wavelength-calibrated 1-D spectra from the reduced 2-D CCD images. As a first task, the object spectra with a corresponding accurate background estimation have to be extracted from the CCD image. For each object an extraction aperture with two adjacent background windows is defined by viewing a CDD cut along the spatial direction for every mask slit. These object apertures are traced along the dispersion axis by following the peak of the spatial source profile. The object flux in the defined apertures is then summed in the spatial direction to recover the full source signal. The same procedure is applied to the designated background pixels, where the sky contribution is scaled to the same area per pixel element as the object aperture, and then subtracted to yield raw 1-D spectra with source signal only. Due to the dominating sky emission lines in the red part of the spectrum (right panel of Fig. 8.1), this last sky subtraction step is critical for the resulting signal-to-noise ratio. The extracted 1-D spectra for each object contain the relative source flux values as a function of image coordinate in the dispersion direction. The next task is to find the exact transformation of the image coordinates to physical wavelength units, i.e. to determine the dispersion solution. This is achieved with the help of wavelength calibration observations using an internal Helium-Argon lamp as a source with numerous sharp and well known emission lines spread over the whole spectral range. The calibration lamp data have been taken with the same slit configuration and ideally at the same telescope position as the science target and are reduced in the same way as the target spectra. For the last extraction step, the 1-D calibration spectra are obtained from exactly the same aperture traces along the detector as the science objects, which results in one raw image-coordinate- matched Helium-Argon calibration spectrum for each object. By identifying several dozen calibration lines with their physical wavelength over the full spectral range, the dispersion solution is derived for each spectrum by iteratively fitting a smooth polynomial function to the data points. The final converged functional form for transforming image coordinates to wavelength is now applied to the raw science spectra yielding the final wavelength calibration. After some manual cleaning of sky line residuals, the calibrated 1-D science spectra are ready for the redshift determination of the next section.

8.4 Redshift Determination

In principle, the identification and subsequent wavelength measurement of a single spectral feature determines uniquely the galaxy redshift according to Equ. 8.4. However, for routine survey applications it is favorable to use a more elaborate cross-correlation analysis for several reasons: (i) for passive galaxies the continuum shape and relative position of the spectral features is important; (ii) all spectral absorption features are taken into account and thus a higher accuracy is achieved; (iii) the quality and error of the redshift measurement can be quantified; (iv) the procedure can be fully automated. The cross-correlation analysis (Tonry and Davis, 1979) uses template spectra with high signal-to-noise ratios as a reference. Since our main interest focuses on absorption line spectra of passive galaxies, good templates for a representative selection of different subclasses of elliptical and lenticular galaxies are of prime importance as a correlation input. 8.4. REDSHIFT DETERMINATION 171

The object spectra for the redshift determination need to be prepared for the correlation analysis in three ways. First, emission lines are removed by replacing the spectral sections deviating more than a given threshold above a smooth continuum fit. This measure is performed since emission lines could originate from physically different locations in a galaxy (e.g. H II regions in spiral disks, and galactic winds) compared to the old stellar populations responsible for the absorption features, and could thus have a slightly different intrinsic redshift. Second, atmospheric spectral features need to be excised or removed to avoid an artificial correlation biasing, in particular with the strong telluric absorption lines (left panel of Fig. 8.1). Third, the heliocentric velocity correction is to be computed for each observation. With this correction, the stationary reference point for velocity comparisons is placed at the solar system barycenter, i.e. the center-of-mass, and takes out the seasonal and nightly velocity modulations of up to 30 km s−1 due to Earth’s motion relative to this reference point. ± The input spectra are now ready for the xcsao correlation analysis which performs the following simplified steps (Kurtz and Mink, 1998): (i) Object and template spectra are rebinned to a logarithmic wavelength scale in order to achieve redshift independent separations between spectral features. From Equ. 8.4 we arrive at ln(λobs)=ln(1+z)+ln(λ0) and thus at z-independent logarithmic intervals d ln(λobs)= d ln(λ0). (ii) The spectral continuum is fitted by a smooth function and subsequently subtracted to extract the absorption lines. (iii) The rebinned, continuum-subtracted template and object spectra are Fourier transformed into frequency space, where they are filtered for high and low frequencies beyond the instrumental resolution limits. (iv) The cross-correlation function in wavelength space cλ(z) between object spectrum oλ and all template spectra tλ is proportional to the convolution of the spectra c (z) o t (z), i.e. the correlation signal obtained when λ ∝ λ λ shifting the template spectrum to redshiftN z (with logarithmic binning). In practice, the correlation function is computed in Fourier space, where the convolution simplifies to a multiplication of the object spectrum with the template spectra. (v) The location of the highest correlation peak in any of the template correlation functions is interpreted as the proper redshift of the observed spectrum. (vi) From the width of the peak a velocity error estimate is obtained, whereas the peak height relative to the average noise peaks yields a measure for the quality and reliability of the determined redshift. After cross-correlation redshifts have been obtained for all available object spectra, the galaxy cluster redshift is determined as the outlier-clipped average of all cluster members. An object is classified as a cluster galaxy if its redshift deviates less than 3–4 times the expected velocity dispersion from the median redshift, i.e. within typical member selection cuts of about 3000kms−1 relative to the average cluster redshift. ± As a last application of this chapter, the proper measurement of the cluster radial velocity dispersion is discussed. Following Danese, De Zotti, & Tullio (1980), the spectroscopic cluster galaxy redshifts can be decomposed in three components: (i) the peculiar velocity of the Milky Way z0, (ii) the proper cosmological redshift zR due to the Hubble expansion, and (iii) the galaxy peculiar velocity in the cluster zG. These components have 5 to be multiplied (not summed) to yield the observed object redshift zobs according to

5The multiplication is intuitive when considering that the cluster velocity dispersion interval is stretched by the cosmological redshift by a factor (1+zR) as all intervals of observables. 172 CHAPTER 8. SPECTROSCOPIC FOLLOW-UP

zobs = (1+ z0)(1 + zR)(1 + zG) . (8.5) Neglecting the ﬁrst term z as a (10−3) eﬀect and approximating the proper cosmological 0 O redshift by the average cluster redshift zR z¯, we are left with the contribution of the galaxy motion within the cluster. This redshift≈ component is to be interpreted as a special relativistic Doppler shift due to the line-of-sight velocity vk of a particular galaxy in the cluster potential

2 ∆λ c + vk vk vk 1+ zG =1+ = 1+ + 2 . (8.6) λ0 sc vk ≈ c O c ! − Since zG 1 for the Doppler shifts of galaxies in clusters, only the ﬁrst order term needs to be taken≪ into account for the last approximation. Plugging this result into Equ. 8.4 and solving for the line-of-sight velocity vk of a given galaxy yields

zobs z¯ vk = c − . (8.7) 1+¯z The ﬁnal result for the velocity dispersion of a galaxy cluster with N spectroscopic member galaxies with measured redshifts zobs(i) is then

2 N 2 2 2 c zobs(i) z δ σk = − , (8.8) N 1  1+ z !  − 1+ z ! − Xi=1 where the last term corrects for the spurious contribution due to the intrinsic average redshift measurement error δ. For a meaningful estimate of the true velocity dispersion σk a minimum of about ten cluster member redshifts is required. We have now accomplished the main objectives of step 5 for the spectroscopic conﬁrma- tion by measuring an accurate cluster redshift, determining cluster galaxy memberships, and deriving a velocity dispersion estimate. This also completes the technical part of Chaps. 6–8 with discussions on the applied X-ray, near-infrared, and optical observational methods and strategies for the XDCP survey.

8.5 Summary

The main spectroscopy contributions within this thesis work include the following: Reduction and analysis of two NORAS 2 spectroscopy runs (see Fig.4.5) with data • of 35 low redshift clusters (results not included); Reduction and spectroscopic analysis of a z 1 XDCP cluster candidate obtained • ∼ through DDT time (see Sect. 10.2); Led a spectroscopy proposal for the GMOS instrument at the 8 m Gemini South • telescope with 16 hours of class C time awarded; observations not executed; PI of an experimental NIR spectroscopy DDT proposal with Gemini’s GNIRS in- • strument for a high-redshift cluster conﬁrmation; not accepted. Chapter 9

XDCP Survey Status and Outlook

This chapter summarizes the current XDCP status and provides cluster candidate and survey characteristics. In addition, open survey tasks will be brieﬂy discussed.

9.1 X-ray Analysis Status

The complete detection and identiﬁcation procedure of serendipitous extended distant cluster candidate sources for the current XDCP survey phase has been discussed in detail in Chap. 6. In this section, a preliminary assessment of the global survey characteristics is provided.

9.1.1 Sky coverage An essential ingredient for the evaluation of the survey search volume (Equ. 4.2) is the computation of the survey sky coverage as a function of limiting flux. The final determination of the XDCP selection function, i.e. the probability to find a cluster with properties C = [z, LX, TX,rc, M, ...] in the given data set with local characteristics D = [teff , B, Θ, ...], is a challenging task for XMM-Newton surveys and will require extensive simulations over the next two years (see Sect. 9.1.3). At this point, approximate global XDCP characteristics can be derived based on some simplifying assumptions. 1 The dashed lines in Fig. 9.1 show the cumulative sky coverage Ssky(

1 Ssky(

Figure 9.1: XDCP sky coverage. Dashed lines indicate the full survey area (blue), 79.3 square degrees in total, and the coverage with a 12′ off-axis angle restriction (black), 50.6 square degrees in total, as a function of limiting point source sensitivity in the 0.5–2.0keV band. Solid lines show conservatively estimated extended source sensitivities corresponding to approximately a 4 σ extent significance and 100 source counts in the single 0.35–2.4 keV detection band. At these levels, to be interpreted as upper limits, the median sensitivity for the full area is 1.7 10−14 erg s−1cm−2 (dotted × red line), and 1.4 10−14 erg s−1cm−2 for the restricted core survey region (dashed red line). × the effective exposure time teff at the pixel position, and (iii) the PSF dependent detection aperture at the given off-axis angle. The determined local point source sensitivities for each pixel element are then merged for the 469 XDCP survey fields. Sky areas with XMM coverage in multiple observations are corrected for by only retaining the most sensitive value of all overlapping pixels. The resulting exact total sky coverage for the XDCP survey ′ is Ssky(tot)=79.3 square degrees for the full area and Ssky(Θ 12 )=50.6 square degrees for the restricted core survey region. The point source sensitivity≤ varies (10–90% −15 −1 −2 of Ssky) within the full survey (blue) by about a factor of five from 2–10 10 erg s cm , and by about a factor of 3.5 from 2–7 10−15 erg s−1cm−2 for the core survey× region (black). × Conservative preliminary estimates of the effective XDCP sky coverage as a function of the extended source sensitivity can be obtained by applying an empirically determined offset factor to the point source sensitivity. At the detection limit, the average point source contains a total of about 25 counts in the 0.35–2.4 keV band. Comparing this value to 9.1. X-RAY ANALYSIS STATUS 175

the identified extended sources in Fig. 6.20 and choosing a conservative extended source count number limit of 100 photons, corresponding to about a 4 σ extent significance, ∼ an average offset factor between the point source detection limit and a realistic reliable extended source threshold of a factor of four is obtained. Applying this empirical offset to the point source sensitivity yields a conservative XDCP sky coverage estimate for extended sources illustrated by the solid lines in Fig. 9.1. The median extended source sensitivity for the full XDCP survey is then approximately f med(tot) 1.7 10−14 erg s−1cm−2 (dotted red line) and f med(Θ 12′) X ≃ × X ≤ ≃ 1.4 10−14 erg s−1cm−2 (dashed red line) for the core survey region. The median XDCP survey× sensitivities determined in this way are to be considered as conservative upper limits since the empirical offset factor of four compared to the point source threshold is quite large. The question how far the extended source sky coverage curves (solid lines) can be shifted to the left, i.e. towards lower limiting flux levels, while retaining a high completeness will be answered with simulations in the near future.

9.1.2 Cluster number counts The cumulative raw XDCP cluster number counts in the 0.5–2.0 keV band for distant cluster candidates (top blue line) and all identified cluster candidate sources (top green line) are shown in Fig. 9.2, where the lighter lines indicate the distribution for a core survey restriction with a maximum off-axis angle of 12′. This log N–log S (see Equ. 3.45) is to be considered as raw in the following sense: (i) the number counts are not normalized to a unit solid angle, (ii) the candidate sample still contains false positives, (iii) the flux determination is based on fiducial values for the Energy Conversion Factor (see Sect. 6.3.3) and the β parameter (Equ. 2.6), and (iv) fluxes are consistently determined for the PN detector, which is not available for about 10% of the (missing) candidate sources due to a non-imaging operation mode of the PN for this fraction of all observations. The overall shape of the cumulative number count distribution yields important diagnostic information for the survey sensitivity characteristics. The flux level where the log N–log S slope starts deviating from the observed number counts in deeper surveys (see Fig. 4.6) indicates the approximate flux limit of the survey below which the source counts become incomplete. In Fig. 9.2 this estimated XDCP completeness level is represented by com −14 −1 −2 the red dashed line at fX 1 10 erg s cm . The log N–log S approaches a constant value, i.e. no additional≃ sources× are found at lower flux levels, at the point where the effective sky coverage of the survey at the given limit drops rapidly towards zero. This approximate minimum survey sensitivity is indicated by the second dashed line at min −15 −1 −2 fX 5 10 erg s cm . At the bright end, the observed cumulative number count distribution≃ × appears to be (artificially) truncated for the most luminous clusters. This effect can be attributed to two features of the source detection algorithm which (i) has a fixed upper size limit of 80′′ for the core radius in the XDCP pipeline setup and (ii) the intrinsic tendency of the eboxdetect task to split very extended cluster sources into several sub-clumps. 176 CHAPTER 9. XDCP SURVEY STATUS AND OUTLOOK

Figure 9.2: Raw log N–log S of XDCP candidates and identified clusters. The total number of objects is shown as a function of the 0.5–2.0 keV flux measured for the PN detector, which is available for about 90% of all sources. The green lines display all identified clusters and distant candidates for the full XMM detector area (top green line) and the FoV restricted to 12′ off-axis angle (lower green line). The blue lines illustrate the cumulative distribution of the distant cluster candidates only. Vertical red lines indicate the position where the log N–log S curves start to turn over at approximately 1 10−14 erg s−1cm−2and the flux level where the effective area becomes small × around 5 10−15 erg s−1cm−2. ×

With the total number of cluster candidates and the survey area available, we can now estimate the observed XDCP galaxy cluster surface density. In total, the full XDCP sample contains 990 unique cluster candidates, 276 of which are distant cluster candidates and 714 are DSS-identified cluster sources. For the restricted core survey at off-axis angles of less than 12′, the 226 distant and 526 DSS-identified candidates make up a combined core sample of 752 sources. Subtracting the approximate false positive fraction of 1/3 for distant and 1/6 for low redshift candidates, the expected sample size of real galaxy clusters is obtained as 780 for the full region of 79.3 square degrees and 590 objects for the core survey area∼ of 50.6 square degrees. Note that this cluster sample∼ size is comparable to the largest local all-sky surveys (see Sect. 4.2), with the difference that the XDCP survey is not aiming at a full identification of all clusters. The approximate average XDCP surface density is thus 9.8 clusters per square degree ∼ for the overall area and 11.6 clusters per square degree for the more sensitive core survey area with Θ 12′. From∼ the observed log N–log S of Fig. 4.6, a cluster surface density of 10 objects per square≤ degree is obtained at a flux limit of 1 10−14 erg s−1cm−2. The identified × combined XDCP cluster candidate sample is thus fully consistent with an average limiting 9.1. X-RAY ANALYSIS STATUS 177

Figure 9.3: XDCP distant cluster candidates in the 0.5-2.0keV flux versus core radius plane. The parameter space of detected candidates is confined by the XMM resolution limit (vertical red line) at core radii of about 6′′ and the background limit (lower red line), where the cluster surface brightness drops below the detection threshold. Green diamonds indicate the positions of three most distant spectroscopically confirmed clusters in the Southern hemisphere (using the source parameters of the XDCP X-ray pipeline), from left to right: XMMU J2235-25 at z = 1.39, XCS J2215-17 at z = 1.45, and RDCS J1252-29 at z =1.24, see Sect. 9.5 for references.

¯lim −14 −1 −2 sensitivity of fX 1 10 erg s cm for the full survey area and slightly below this value for the core≃ sample.× The estimated XDCP sensitivity limits from the raw log N–log S and the total candidate numbers in this section are about 30% lower than the conservative median sensitivities derived in the previous section. This apparent discrepancy originates from the attempt to lim characterize an XMM serendipitous surveys with a single number, the flux limit fX . The difficulties arising from this concept of a global survey flux limit will be briefly discussed in the next sub-section.

9.1.3 Selection function In this final section on X-ray properties of XDCP cluster candidates, some important aspects concerning the selection efficiency of serendipitous extended sources with XMM- Newton are summarized. Figure 9.3 displays the extended X-ray sources of the XDCP distant cluster candidate sample in the 0.5-2.0 keV flux versus core radius3 plane (see Equ. 2.6), the two prime source

3 Core radii rc are determined with emldetect (see Sect. 6.3.3), i.e. using a ﬁxed β =2/3. 178 CHAPTER 9. XDCP SURVEY STATUS AND OUTLOOK

parameters governing the detection probability. This empirical plot illustrates nicely the two main XMM limitations that constrain the accessible parameter space of distant cluster candidate sources (see Sect. 6.1). The XMM resolution limit sets a lower core radius cut-off at about 6′′ (vertical red line), whereas the background imposes a limit on the maximum core radius for sources of given flux (lower red line). Note that the currently three most distant spectroscopically confirmed clusters in the Southern hemisphere (green diamonds) are well away from the sample limits, in particular concerning the accessible minimum flux threshold (vertical offset to the lower red line). The selection function characterizes a galaxy cluster survey in terms of the extended source detection probability of clusters with given properties in the X-ray data. The most important cluster characteristics with prime influence on the detection rate are given in the following:

Source Flux: For an idealized, homogenous X-ray survey, the flux limit is the single characteristic that determines the detection probability of a source. The received X- ray flux fX in a given energy band depends on the cluster luminosity LX, the spectral shape governed by the gas temperature TX, the redshift z, and the galactic hydrogen absorption column NH. Note that high cluster redshifts lead to severe cosmological surface brightness dimming (Equ. 3.22) and can hence introduce selection biases at faint flux levels, where the surface brightness drops below the signal-to-noise threshold (for a real survey with background).

Core Radius: Once the background surface brightness reaches levels comparable to that of the source, the physical size of the cluster, characterized by the projected core 4 radius rc (see Equ. 2.6), becomes the second prime property governing the detection rate. For further analytical considerations it is important to note that the source ﬂux fraction encircled within the core radius fX(r

Cooling Core Activity: Cooling core activity in clusters (see Sect. 2.3.5) is connected to very peaked inner SB profiles with smaller core radii than non-cooling core clusters (e.g. Chen et al., 2007). For possible high-redshift CCC, the peaked inner profile can have positive or negative effects on the detection probability. Moderate cooling core activity, as found for XMMU J2235.3 2557 at z =1.393 (Santos et al., submitted), increases the SB contrast and hence improves− the signal-to-noise ratio. However, if the core radius of a strong CCC drops below the XMM resolution limit, the extended source detection probability drops and the cluster might be missed.

4 To avoid confusion with the oﬀ-axis angle Θ, the notation rc is used for the projected core radius throughout the following discussion. 5This can be shown by integrating Equ. 2.6. 9.1. X-RAY ANALYSIS STATUS 179

Cluster Morphology: For the detection process, cluster sources are usually idealized as spherically symmetric. This might not be a good approximation (e.g. Hashimoto et al., 2005), in particular at high redshifts, where clusters are either dynamically very young objects or are still in the formation process with a high merger frequency (see Fig. 3.4). Deviations from spherical symmetry and in particular cluster substructure will lower the detection probability, since the cluster emission might be mistaken for several point sources.

AGN Activity: If a cluster hosts an X-ray bright AGN, the chance for a misidentification as a point source increases. While only 5% of local clusters have bright AGN close to their center (Böhringer et al., 2004), the point source contamination for high-redshift systems can be significant (e.g. Stanford et al., 2001).

The relevant XMM instrumental and data characteristics for the detection of faint extended sources have been discussed in Sect. 6.1. As a function of detector position, i.e. off-axis angle Θ and azimuth φ, the key features are (i) the effective, vignetting corrected clean exposure time teff (Θ), (ii) the PSF FWHM(Θ) defining the minimum resolvable cluster angular size rmin(Θ), (iii) the background surface brightness B(Θ, φ), and (iv) detector defects (e.g. gaps, dead columns, hot pixels). Close to the detection threshold for extended sources, XMM observations are, in very good approximation, always background limited (see Equ. 7.1). Assuming that the detection threshold is defined by a minimum signal-to-noise ratio SNRmin in a given detection aperture, e.g. the core radius of the source, then the limiting flux threshold scales as f B(Θ,φ)/ t (Θ) with the exposure time t and the background level B. Sim- lim ∝ eff eff ilarly,q the increasedq detection aperture of larger extended sources at a given background level has the scaling effect f πBr2 r . This expected linear dependance of the core lim ∝ c ∝ c radius on the detection limit is reflectedq by the lower red line in Fig. 9.3. A priori unknown effects can be taken into account by introducing fudge factors for the cluster AGN and foreground flux contamination of nearby sources Acontam > 1 and for possible morphological ∼ cluster distortions Bmorph > 1. By defining the local flux limit as the level where a cluster ∼ with resolvable core radius rc >rmin(Θ) is detected with a specified probability, e.g. 50%, then the combination of all factors yields

B(Θ,φ) flim(rc >rmin) rc q Acontam Bmorph = const . (9.1) ∝ teff (Θ) 6 q The effective local extended source flux limit is hence a function of the XMM field characteristics (teff , B), the location on the detector (Θ, φ), the sky position for contaminating nearby sources Acontam(RA, DEC), and the cluster properties (rc, Bmorph). A single global flux limit for background limited XMM serendipitous surveys is hence not well defined. The variation of the cluster core radius over typical values in the local Universe of 50-500 kpc already introduces a difference in the limiting sensitivity of one order of magnitude. A stated flux limit for XMM surveys is only meaningful for a specified cluster size, e.g. for 180 CHAPTER 9. XDCP SURVEY STATUS AND OUTLOOK

′′ core radii of rc =15 , which can be resolved over the full FoV. All other factors combined cause sensitivity variations across the XDCP survey of approximately another factor of ten from the deepest to the shallowest fields. The currently best modelled selection function of all XMM-Newton surveys has been provided by the XMM-LSS project (see Sect. 4.3). For the published five square degrees of data with 29 clusters, Pacaud et al. (2006, 2007) have determined the detection probability of sources as a function of core radius from simulated data assuming symmetric β-model cluster profiles. As a result, the XMM-LSS survey has abandoned the concept of a single global flux limit, for the discussed reasons, and replaced it by local likelihood thresholds for the source extent and detection significance. In close collaboration with a team led by Joe Mohr at the University of Illinois, simulation efforts have started to determine the XDCP selection function for each of the 469 survey fields. The final goal is to go one step beyond the XMM-LSS modelling and use realistic artificial clusters from high-resolution simulations with different masses, redshifts, morphologies, and dynamical states. These simulated clusters will be placed in mock images that take the XMM-Newton instrumental characteristics into account. After adding background and a point source population that follows the observed AGN number counts, the XDCP detection pipeline will be applied to multiple mock realizations of the same field. From the comparison of the detected extended source list with the cluster input catalog, the completeness and contamination functions can be evaluated for each field as a function of input cluster mass M, size rc, morphology, and redshift z.

9.2 Imaging Status

The results of the two Calar Alto Z–H follow-up imaging campaigns (Chap. 7) are summarized in Fig. 9.4. In total, data was obtained for 63 distant cluster candidates. The redshift distribution for 43 systems with detected optical counterparts is shown in the top panel, based on Z–H red-sequence redshift estimates for most of the sources and some additional available redshifts from the literature (see Sects. 10.1 & 10.2). The red hashed region displays secure cluster detections, the top dashed line also includes systems with lower confirmation confidence, usually attributed to sparse, less obvious red-sequences (if real). Two sources with central galaxy overdensities exhibit an ambiguous color distribution and are not included in the plot. Three objects with counterparts in the H-band still lack Z-band data for a redshift estimate, and 15 X-ray sources were classified as spurious. In combination with a few additional expected non-cluster sources from the lower confirmation confidence sample, an approximate false positive fraction of 1/3 is obtained, as was already discussed in Sect. 6.4. A total of 21 cluster candidates with photometric redshift estimates of z>0.9 were discovered, the 15 secure systems being high-priority candidates for spectroscopic follow-up. The center panel of Fig. 9.4 displays the redshift histogram of DSS-identified low redshift cluster candidates (see Sect. 6.4) that could be imaged together with distant targets at no extra observational cost. As discussed in Sect. 6.4.3, the blue histogram confirms 9.2. IMAGING STATUS 181

that a good fraction of clusters out to z 0.6 can be successfully identified during the classification procedure, with the positive∼ consequence that the resulting distant cluster candidate distribution (red histogram) is skewed towards a peak at z 0.9. The large fraction of uncertain identifications at low redshifts (top blue line) is to∼ be attributed to low mass groups with only few probed member galaxies and expected Z–H colors in the CMD regions associated with high background object densities. The lower black panel illustrates the combined Calar Alto cluster sample obtained from the sum of the upper red and blue distributions and five additional XMM target clusters that are not part of the XDCP serendipitous cluster catalog. The observed distant cluster candidates at Calar Alto constitute about 1/4 of the complete XDCP sample. XDCP imaging data for about 2/3 of all candidates have been obtained and will likely be extended to 80% by the end of 2007. In total, the XMM-Newton Distant Cluster Project has been awarded with ten follow-up imaging runs at four different observatories making use of four different cluster identification techniques (see Sect. 7.1):

The pilot study (Sect. 5.5) started with a two semester VLT FORS2 R–Z imaging • program in the years 2003/2004;

Additional FORS 2 R–Z imaging data of cluster candidates was obtained during two • recent runs in 2006/2007;

The two Calar Alto Z–H campaigns in 2006 have been discussed in Chap. 7, results • will be shown in Chap. 10;

The I–H method has been tested during two runs in 2007 using the SOFI and EMMI • instruments at ESO’s New Technology Telescope (NTT);

A diﬀerent multi-band photometric redshift method using g r i z imaging complemented • by H-band observations for very distant cluster candidates has been pursued at the CTIO observatory in 2006 and will be continued in fall 2007.

The latter multi-band imaging program is being conducted in close collaboration with the group of Joe Mohr at the University of Illinois. The scientific aim of this special XDCP sub-project is the optical identification and characterization of all X-ray selected clusters in the SZE survey region of the South Pole Telescope (SPT) (see Sects. 2.3.6 & 12.1). The future SZE coverage of the 106 fully analyzed XDCP fields in the SPT region (red symbols in Fig. 6.5) with a total effective clean XMM exposure time of almost 2 Msec will open new possibilities for joint X-ray-optical-SZE cluster studies. The X-ray selected sample of approximately 200 clusters with multi-band photometric redshifts over a large baseline will allow a detailed survey cross-calibration between the X-ray-based XDCP and the SZE survey of the South Pole Telescope, a critical pre-requisite for the use of distant clusters as sensitive cosmological probes. 182 CHAPTER 9. XDCP SURVEY STATUS AND OUTLOOK

Figure 9.4: Histogram of the final Calar Alto imaging cluster sample based on estimated Z–H red- sequence redshifts. Top: Redshift distribution of the targeted distant cluster candidates in ∆z = 0.2 bins with a secure photometric confirmation (red hashed) and including sources with lower photometric confirmation confidence (top line). Vertical dashed lines indicate redshifts of 0.5 and 1.0. Center: Distribution of DSS-identified cluster candidates that were included in the OMEGA2000 field-of-view of distant candidates. Bottom: Redshift histogram of the combined Calar Alto X-ray cluster sample as the sum of the two upper red and blue curves and five additional XMM target clusters, which are not part of the serendipitous object class. The total Calar Alto sample includes 21 X-ray cluster candidates with photometric red-sequence redshift estimates of z>0.9, 15 of them with a high confidence level. 9.3. SPECTROSCOPY STATUS 183

9.3 Spectroscopy Status

The spectroscopic conﬁrmation of z > 1 galaxy cluster candidates is the observational bottleneck of the XDCP survey since it requires the highly competitive use of red-sensitive multi-object spectrographs on 8 m-class telescopes (see Chap. 8). The XDCP status of spectroscopically conﬁrmed high-redshift clusters can be summarized as follows:

XMMUJ2235.3-2557 at z = 1.393 was the ﬁrst spectroscopically targeted XDCP • distant cluster in 2004 (Sect. 5.4);

XMMU J0104.4-0630 was the second followed-up pilot study candidate in 2005 and • was conﬁrmed at z =0.947 (see Sect. 10.2);

Three cluster candidates with estimated redshifts of z 0.9 have been spectroscopi- • ∼ cally observed at one of the 6 m-class Magellan telescopes;

Four other XDCP distant cluster candidates have been spectroscopically conﬁrmed • through other channels and surveys (see Sect. 9.5) at redshifts of 1.45, 0.976, and two at z =1.05;

Two spectroscopic VLT FORS 2 XDCP programs are currently being executed tar- • geting a dozen high-redshift objects, completing the pilot study program on one hand and starting the conﬁrmation of Z–H selected Calar Alto candidates on the other;

Spectroscopy time for eight additional FORS 2 observations has been granted for the • upcoming semester.

9.4 XDCP Outlook

With the X-ray candidate selection for the current survey phase completed (Chap. 6) and more than half of all XDCP distant cluster candidates followed-up with imaging observations (Chap. 7), the XDCP project focus will shift towards the spectroscopic confirmation of high-redshift clusters (Chap. 8) over the next few years. The XDCP X-ray analysis will change its scope from the search and identification of distant cluster candidates to the detailed characterization of the survey sample. The initiated collaborative project of realistic simulated cluster data (Sect. 9.1.3) will be used to evaluate the XDCP selection function in detail. By applying the XDCP X-ray pipeline to simulated cluster fields, the probability for the detection of extended sources as a function of various input parameters (e.g. off-axis angle, exposure time, core radius, background) will be quantified in conjunction with the fraction of missed cluster sources (e.g. due to AGN dominated X-ray flux, distorted morphologies, compact cores). With this information at hand, the final strict X-ray selection cuts for the XDCP core survey cluster sample will be defined, which will provide the assessment of the associated cluster search volumes for the upcoming cosmological studies. 184 CHAPTER 9. XDCP SURVEY STATUS AND OUTLOOK

After a full evaluation of all available follow-up imaging data of distant cluster candidates and the different identification techniques, the completion of the final quarter of the two-band photometric identification program of XDCP can be prepared in 2008. The available (and scheduled) optical coverage of the full XMM field-of-view for approximately 20% of the survey overlapping with the SPT region will additionally allow the quantitative assessment of the applied screening procedure of Sect. 6.4, in particular the fraction of false negatives, i.e. detected real distant cluster sources that were not added to the distant cluster candidate list. The completion of the spectroscopic follow-up has the longest time line and will likely require 3–4 more years. The compilation of an X-ray selected, well characterized and spectroscopically confirmed z >1 cluster sample over 50 square degrees within this time span will be highly competitive∼ as a prime cosmological∼ tool and as pathfinder sample for several of the upcoming larger cluster surveys (e.g. SPT, eROSITA, see Chap. 12). In principle, the XMM-Newton Distant Cluster Project could be extended in the future on several different scales, provided sufficient man power is available. First, the planned enlargement of the XDCP sky coverage in the SPT region will add approximately 100 additional XMM archive fields to the survey that have accumulated over the past three years. Second, new archival data could be included for the full Southern hemisphere XDCP sample based on the selection criteria discussed in Sect. 6.2, which would increase the number of XMM fields by about 400. The largest sky coverage increase with more than 600 new XMM fields would result from a Northern sky (DEC > +20◦) extension of the survey. Suitable efficient follow-up imaging methods applicable for Northern sky observatories have been developed and tested (Chap. 7) and new large facilities, such as the Large Binocular Telescope (LBT), could be involved in the spectroscopic distant cluster work. In total, XMM archival data for a tripling of the current XDCP survey sky coverage up to 150 square degrees has accumulated, but the related work to fully exploit this ∼ unique data set would fill several more PhD theses in the future. The highest priority of the XMM-Newton Distant Cluster Project is therefore the completion of the ongoing survey phase and the scientific exploitation of the first X-ray selected distant cluster sample with >30 systems at redshifts of z>1. The top three XDCP science drivers to be addressed∼ over the coming years are (i) the determination∼ of the cluster number density evolution out to z 1.5 for constraints of prime cosmological parameters (see Sect. 3.3), (ii) the characterization∼ of evolving cluster properties at z > 1 as a function of redshift and cluster mass (see Sects. 2.3.3 & 2.3.4), and (iii) the exploration∼ of the cluster galaxy population evolution at large lookback times (see Sect. 2.4). First XDCP results on galaxy evolution in clusters will be presented in Chap. 10. 9.5. SPECTROSCOPICALLY CONFIRMED Z>1 CLUSTERS IN 2007 185

ID Object Name Survey z fX,0.5−2.0 keV LX,bol TX erg s−1cm−2 1044 erg s−1 keV 1* XMMXCS J2215.9 1738 XCS 1.45 1.2 10−14 4.4 7.4 − × 2* XMMU J2235.3 2557 XDCP 1.393 2.6 10−14 11.1 6.0 3 CIG J0848.6+4453− RDCS 1.273 1.0 × 10−14 1.0 2.9 × 4 RX J0848.9+4453 RDCS 1.261 1.8 10−14 2.8 5.2 5* RX J1252.9 2927 RDCS 1.237 2.5 × 10−14 6.6 6.0 6 XLSS J0223.1− 0436 XMMLSS 1.22 6.2 × 10−15 1.1 3.8 − × 7 RX J1053.7+5735 LH 1.134 2.0 10−14 2.0 4.9 8 RX J0910.8+5422 RDCS 1.106 1.1 × 10−14 2.5 7.2 × 9* XLSS J0224.1 0413 XMMLSS 1.05 3.1 10−14 4.8 4.1 10* XLSS J0227.1−0418 XMMLSS 1.05 1.1 × 10−14 1.5 3.7 11 Cl J1415.1+3612− WARPS 1.03 1.1 × 10−13 10.4 5.7 × Table 9.1: Spectroscopically confirmed z > 1 X-ray luminous galaxy clusters in August 2007. Specified values for the X-ray flux fX, bolometric luminosity LX, and temperature TX are rounded, see references in text for details. Marked objects (*) in the first column are part of the XDCP distant cluster sample.

9.5 Spectroscopically Conﬁrmed z >1 Clusters in 2007

This chapter on the XDCP survey status is closed with a compilation of all spectroscopically confirmed X-ray luminous galaxy clusters at z> 1 published by August 2007. Since the beginning of this thesis work the total number of z>1 X-ray clusters has increased from five to eleven, indicating the advances in the XMM-Newton era on one side but also emphasizing that these systems are still ‘rare beasts’. Table 9.1 lists the cluster names, surveys within they were confirmed, the redshift, and basic cluster properties of the eleven known systems. Five out of six confirmed z > 1 clusters in the Southern hemisphere are part of the XDCP distant cluster survey sample indicated by the star in the first column6. Note that several optically selected clusters at high redshift have recently been spectroscopically confirmed, most notable two infrared selected clusters with Spitzer at redshift z =1.41 (Stanford et al., 2005) and at z =1.51 (McCarthy et al., 2007). These clusters are not yet confirmed in X-rays and are therefore not listed in Tab. 9.1. The references for clusters with identification number [ID] are: [1] Stanford et al. (2006), [2] Mullis et al. (2005), [3] Stanford et al. (1997), Ettori et al. (2004), [4] Rosati et al. (1999), [5] Rosati et al. (2004), [6] Bremer et al. (2006), [7] Hashimoto et al. (2005), Hashimoto et al. (2002), [8] Stanford et al. (2002), [9,10] Andreon et al. (2005), Pierre et al. (2006), and [11] Maughan et al. (2006), Perlman et al. (2002).

6RXJ1252.9-2927 was also detected as a serendipitous source in addition to the targeted observations. 186 CHAPTER 9. XDCP SURVEY STATUS AND OUTLOOK Chapter 10

High-Redshift Cluster Studies: First Results

The results of the follow-up observations are now used to address some of the scientific questions discussed in the introductory Chaps. 2 & 3. The cluster studies in Chaps. 10 & 11 rely on photometric data obtained during two imaging campaigns at the Calar Alto Ob- servatory (see Tab. 7.5 and Chap. 7). This chapter will focus on early results based on the available sub-sample of spectroscopically confirmed clusters. In Sect. 10.1, the new Z–H method is calibrated with the spectroscopic sample in order to establish a reference model for the red-sequence color evolution and to provide first constraints on the formation epoch of the bulk of the stellar populations in early-type galaxies. This model is the basis for the photometric redshift estimates of the full distant cluster candidate sample. Section 10.2 presents the spectroscopic, photometric, and X-ray analysis of a newly confirmed galaxy cluster at z =0.95. This example provides a first a posteriori test for the photometric Z–H red-sequence technique and strengthens its applicability to the cluster identification and redshift estimation.

10.1 The Formation Epoch of Red-Sequence Galaxies

Formation models for ellipticals and implications of the observed red-sequence properties of cluster galaxies have been discussed in Sect. 2.4.2. In this section, we compare observed red-sequence Z–H colors of ten calibration clusters with known spectroscopic redshifts to the SSP model expectations (see Sect. 7.1).

10.1.1 Calibration of the Z–H method Figure 10.1 displays the observed evolution of the average Z–H red-sequence color as a function of redshift (blue dots) together with diﬀerent model expectations (solid and dashed lines). In order of increasing redshift, the following clusters have been used to calibrate

187 188 CHAPTER 10. HIGH-REDSHIFT CLUSTER STUDIES: FIRST RESULTS

Figure 10.1: Redshift evolution of the Z–H color of red-sequence galaxies. Measured (average) red-sequence colors with 1 σ uncertainties for ten galaxy clusters with known redshifts in the range 0.18 z 1.45 (blue dots) are shown together with different synthetic evolutionary tracks set by ≤ ≤ different models of passively evolving galaxies (i.e. single burst models). Solid lines represent solar metallicity models with formation redshifts zf =3 (black), zf =5 (blue), and zf =10 (red). The dashed black line indicates the color offset for a zf = 3 model with three times higher metallicity. The blue model (zf =5, Z =Z⊙) has been used to derive red-sequence redshift estimates for all observed cluster candidates. the model for the Z–H color evolution: WARPS J0943+1644 at z =0.180 (Vikhlinin et al., 1998), A383 at z =0.187 and A2390 at z =0.228 (Struble and Rood, 1999), CL0016+1609 at z = 0.541 and RXJ0018+1617 at z = 0.55 (Connolly et al., 1996), XLSSJ0227- 0411 at z =0.581 (Pierre et al., 2006), XMMUJ0104-0630 at z =0.947 (see Sect. 10.2), XMMUJ2215-1740 at z =0.99 0.052 (Lamer et al., in preparation), XLSSJ0227-0418 ± z =1.03 (Valtchanov et al., 2004), and XMMXCSJ2215-1738 at z =1.45 (Stanford et al., 2006).

As can be seen in Fig. 10.1, the solar metallicity zf =5 simple stellar population model

(blue line) is fully consistent with all measurements. For this reason, the (zf =5, Z =Z⊙)-

1An additional redshift error of 0.05 has been added, since the spectroscopic redshift appears uncertain. ± 2Photometric redshift estimate based on deep multi-color CFHT photometry. 3Spectroscopic redshift, but the presence of a second overlapping foreground structure at z 0.9 decreases the reliability (see also Pierre et al., 2006). ∼ 10.1. THE FORMATION EPOCH OF RED-SEQUENCE GALAXIES 189

Figure 10.2: Maximum likelihood fit of the formation redshift zf of passively evolving red-sequence galaxies and their metallicity Z. The long redshift baseline allows simultaneous constraints of both parameters yielding a best fit model (1 σ errors) of z =4.24 1.1 and Z =(1.20 0.35) Z . f ± ± ⊙ model was selected as master calibration reference for the photometric red-sequence redshift estimates of all observed XDCP cluster candidates (see Fig. 9.4). The visual inspection of Fig. 10.1 reveals that the discrimination power between the model formation epochs rests upon the highest redshift cluster at z =1.45. This single data point currently excludes a zf =3 model on the 1.5–2 σ level, whereas the high-redshift zf =10 model is still within the 1 σ uncertainties. Nevertheless, the unprecedented redshift baseline for X-ray cluster red-sequence studies in conjunction with the monotonic color-redshift relation of the new Z–H method allow first reasonable simultaneous constraints of the two main formation model parameters zf and Z. Formation epoch (i.e. age) and metallicity are highly degenerate in terms of resulting colors, although they are independent model parameters (see lower right panel of Fig. 7.1). Whereas the formation redshift impacts the galaxy colors at high redshifts of z> 1, an increasing metallicity results in an almost constant overall reddening of the color. We can hence parameterize the age and metallicity effects as a linear combination of the form

M[5] M[3] M[3 Z ] M[Z ] M [z =(5+A),Z =(1+B) Z )] M[5,Z ]+A − +B ⊙ − ⊙ , (10.1) SSP f ⊙ ≈ ⊙ 2 2 190 CHAPTER 10. HIGH-REDSHIFT CLUSTER STUDIES: FIRST RESULTS

where MSSP[zf ,Z] denotes the simple stellar population model of formation redshift zf and metallicity Z used for predicting the Z–H colors as a function of redshift. The (zf = 5, Z = Z )-model is used as reference, and the formation redshift parameter A z 5 and ⊙ ≡ f − the metallicity oﬀset B (Z/Z⊙) 1 allow continuous model variations based on the linear interpolation of the initially≡ coarse− parameter grid.

10.1.2 Discussion The maximum likelihood fit result of the parameterized model to the data is displayed in Fig. 10.2, where the contours define the 1 σ, 2 σ, and 3 σ confidence regions. The best fit values with 1 σ errors are zf =4.24 1.1 for the formation epoch of the bulk of the stellar populations of red-sequence galaxies± and Z=1.20 0.35 Z for their average metallicity. ± ⊙ The determined direct observational constraint on the formation epoch based on the Z- H color evolution is fully consistent with the lower limits obtained from individual red- sequence studies at high redshift (e.g. Mei et al., 2006a), the age of high-z massive ellipticals (e.g. Dunlop et al., 1996), age measurements of local early type galaxies (e.g. Thomas et al., 2005), and the latest high resolution simulation results (De Lucia et al., 2006). The current constraints are based on only ten clusters with known redshifts. By populating the redshift regime 0.9 < z < 1.5 with the XDCP targets currently scheduled for spectroscopic confirmation, we will∼ soon∼ be able to significantly lower the uncertainties on the cosmic formation epoch of early type red-sequence galaxies in clusters using the Z–H red-sequence method. Eight4 of the calibration clusters over the full redshift baseline together with their color- magnitude diagrams are shown in Figs. 10.3 & 10.4; XMMU J0104-0630 will be presented in Sect. 10.2. The dashed red horizontal lines in the right column indicate the determined red-sequence colors based on the data, dotted red lines illustrate the estimated color uncertainties. Spectroscopically confirmed BCGs are marked by green arrows in images, their corresponding locations in the CMDs are indicated by green circles. Additional confirmed cluster members are circled in blue in Fig. 10.4 for the three calibration clusters with highest redshifts. In preparation of the preliminary BCG study of Sect. 11.1, a closer look at the spectroscopic sample can provide some observational guidelines for the selection and interpretation of BCGs. The BCG identification for all calibration clusters at z < 0.6 is unambiguous based on the central position in the cluster and the total apparent H-band∼ magnitude. At these low and intermediate redshifts, all case study BCGs are located on the red-sequence within the uncertainties of the color determination, which justifies the assumption that stellar populations of BCGs are passively evolving. The reduced redshift sensitivity of the Z–H method at z< 0.9 is partially compensated by the lower color uncertainties of well populated red-sequences. At z > 0.9, the observational task of locating the cluster red-sequence for the redshift ∼ 4The low-mass cluster XLSS J0227-0411 at z =0.58 has the largest uncertainties, i.e. the lowest weight, among the calibration objects at z <0.6 and is therefore not explicitly discussed. ∼ 10.2. INDICATIONS FOR LARGE-SCALE STRUCTURE AT Z 1 191 ∼

estimate is more challenging. The red-sequences are often sparsely populated, as shown in the lower three rows of Fig. 10.4. However, the redshift leverage of the Z–H color increases (see Fig. 10.1) and the background (black small dots) in the CMD decreases for redder colors. In absence of an obvious sequence, the Z–H color of the red envelope (formed by the ﬁlled symbols) promises to be a reliable redshift proxy, as shown for the cases of XLSS J0227-0418 and XMMXCS J2215-1738 in Fig. 10.4. These two case study examples also suggest that BCGs at high redshifts might not sit at the X-ray center and that their color can be slightly bluer than the upper envelope. For the identiﬁcation of BCGs of high-z clusters based on photometric data only, the spectroscopic sample points at an outer radius of 30′′ and maximum color tolerances of 0.5 mag as reasonable search ∼ ∼ constraints.

10.2 Indications for Large-Scale Structure at z 1 ∼ In this section, the full XDCP survey analysis (steps 1–5) is applied to the environment of the X-ray cluster XMMU J0104.4-0630, the second pilot study selected object with available XDCP spectroscopic follow-up data. The relevant background information concerning cluster red-sequences (Sect. 2.4.2) and environmental eﬀects of galaxy evolution (Sect. 2.4.3) have been discussed in the introductory chapter. At the cluster redshift of z 0.95, the lookback time is 7.5 Gyr and one arcsecond corresponds to a physical comoving≃ scale of 7.9 kpc (see Fig. 3.3 & 3.1). The results presented here follow closely the journal letter of Fassbender et al. which has been submitted for publication in Astronomy & Astrophysics.

10.2.1 Observations X-ray Data The galaxy cluster XMMU J0104.4-0630 was selected as a serendipitous extended X-ray source in a high-galactic latitude archival field (b=+69◦; nominal exposure time: 26.7 ksec; OBSID: 0112650401; here field 1). A second, partially overlapping XMM-Newton observation (nominal exposure time: 24.9 ksec; OBSID: 0112650501; here field 2) extends the X-ray coverage to the South, providing moderately deep X-ray data over a field of about 25′ 45′. × The data were reduced following the procedure discussed in Sect. 6.3. High background periods were excluded by applying a two-step flare cleaning procedure first in the hard (10-14 keV) and subsequently in the 0.3-10 keV band following Pratt&Arnaud (2003). The remaining clean exposure times for field 1 (field 2) are 15.1 ksec (13.7) for the EPIC PN camera, 23.2 ksec (19.7) for MOS1, and 23.5 ksec (21.2) for MOS2, resulting in an effective (sensitivity weighted) clean exposure time of about 18 ksec throughout most of the covered field. Images in different bands for the PN and MOS detectors were created from the cleaned event lists and later combined for the overlapping sky region of fields 1 and 2. 192 CHAPTER 10. HIGH-REDSHIFT CLUSTER STUDIES: FIRST RESULTS

Figure 10.3: 2′ 2.′5 image cutouts and Z–H CMDs for the low-z calibration clusters. X-ray × contours are shown in yellow and BCGs are indicated by green arrows (left column) or a green circle (right column). The red-sequence color (red dashed line) and the estimated color uncertainties (red dotted lines) are marked. Dark blue symbols represent objects located within a region of 30′′ from the center, smaller green circles objects within 30–60′′, black dots mark all other objects in the ﬁeld. 10.2. INDICATIONS FOR LARGE-SCALE STRUCTURE AT Z 1 193 ∼

Figure 10.4: 2′ 2.′5 image cutouts and Z–H CMDs for the high-z calibration clusters with known × redshifts. Spectroscopically conﬁrmed cluster galaxies in addition to the BCG are indicated by blue circles for the three most distant systems. In the case of XMMU J2215-1740, galaxies with multi-band photometric redshifts in the range 0.93 z 1.05 are marked. ≤ ≤ 194 CHAPTER 10. HIGH-REDSHIFT CLUSTER STUDIES: FIRST RESULTS

Figure 10.5: Spectroscopic confirmation of XMMU J0104.4-0630. Left: 9′ 9′ H-band image of × the environment of XMMU J0104.4-0630. Spectroscopically confirmed cluster members are indicated by green circles with identification numbers; tentative cluster members are marked by blue circles, radio sources by crosses, and the consistent literature redshift is shown in black. The cluster radio source is located in between the BCG (ID 1) and the galaxy coincident with the X-ray center (ID 2). Right: VLT FORS 2 spectra of the galaxies with secure and concordant redshifts giving the cluster redshift of z =0.947 0.005 as labelled in the left panel (smoothed with a 7pixel boxcar filter). Flux ± units are arbitrary; sky emission lines and telluric absorption lines are shown in red at the bottom and top, respectively. The two spectra from the top belong to the two brightest cluster galaxies in the center, the bottom panel displays the stacked spectrum of the galaxies with IDs 3–7.

A total of six extended X-ray sources were detected in the individual fields using the SAS tasks eboxdetect and emldetect. This work focusses on two of the sources, the main galaxy cluster XMMU J0104.4-0630 (here cluster A) and a second system XMMU J0104.1- 0635 at a projected distance of 6.4′ to the South-West (here cluster B). Cluster A was detected at an off-axis angle of 5.5′ in field 1 with an aperture corrected, unabsorbed flux of −14 −1 −2 20 −2 (1.7 0.3) 10 erg s cm in the 0.5–2.0 keV energy band, using NH =5.08 10 cm from± the 21× cm survey of Dickey & Lockman (1990). The estimated core radius for× a β =2/3 model is (18.8 0.8)′′ corresponding to 150 kpc at z =0.95. The flux of cluster B at 7.2′ off- axis angle in field± 2 is (1.2 0.4) 10−14 erg s−1 cm−2, its estimated core radius of (45 4)′′, corresponding to 340 kpc,± is significantly× larger. The main properties of the systems± are summarized in Tab. 10.1. 10.2. INDICATIONS FOR LARGE-SCALE STRUCTURE AT Z 1 195 ∼

Optical and Near-IR follow-up The sky region around XMMU J0104.4-0630 was followed-up with R (1140s) and Z-band (480s) imaging on 18 October 2003 with VLT-FORS2 (6.′8 6.′8 FoV). The target was × reobserved in H (1000s) and Z (1800s) on 30 October 2006 in photometric conditions (1′′ seeing) using the NIR wide-field camera OMEGA2000 (Bailer-Jones et al., 2000) at the Calar Alto 3.5m telescope with a larger 15.4′ 15.4′ field-of-view, which also covered the × second cluster XMMU J0104.1-0635 to the South-West. The SExtractor photometry in this larger field is calibrated to the Vega system using 2MASS point sources (Cutri et al., 2003) in H, and designated SDSS standard star observations (Smith et al., 2002) in Z (see Chap. 7). In both bands the limiting magnitudes (50% completeness) of Hlim 20.7 and Z 22.8 correspond to m*+1.6 for passively evolving galaxies at the cluster redshi∼ ft (see lim ∼ Tab. 7.2). Figure 10.5 (left) provides an overview of the field, Fig. 10.6 (top panels) displays the R+Z+H false-color composite image of XMMU J0104.4-0630 (left), the Z+H images of XMMU J0104.1-0635 (center), and the region halfway between the clusters (right). The corresponding Z–H color magnitude diagrams are shown in the bottom panels of Fig.10.6. Spectroscopic observations were obtained on 2 November 2005 using a VLT-FORS 2 MXU slit-mask centered on XMMU J0104.4-0630 for a total exposure time of 60 minutes (see Chap. 8). The incomplete execution of the program, originally scheduled for 3 hours, and the rather poor seeing conditions (about 2′′) resulted in a data quality that allowed a redshift determination for only about half the targeted sources. However, seven galaxies were found at the same redshift and could thus be identified as secure cluster members (see Fig. 10.5, right panel), yielding a cluster redshift for XMMUJ0104.4-0630 of z =0.947 0.005. Eight additional galaxies (blue circles in the left panel) are classified as tentative± members, with indications of the D4000 break at the expected position but significant contamination from telluric absorption and sky emission lines.

10.2.2 Discussion Cluster XMMU J0104.4-0630

XMMUJ0104.4-0630 (Fig.10.6, top left) with a 0.5–2.0keV luminosity of Lx =(6.4 1.3) 1043 erg s−1 is an intermediate mass cluster of M 1.1 1014 M and T 3 keV,± based× 500 ∼ × ⊙ X ∼ on scaling relations following Finoguenov et al. (2007). The system has a compact and fairly regular X-ray appearance with slight extensions to the North-East and South-West. The cluster core is dominated by early-type red galaxies as in low redshift clusters. The X-ray center coincides with the second brightest galaxy (ID 2 in left panel of Fig. 10.5). The brightest cluster galaxy and the densest part of the cluster core exhibit an oﬀset of about 10′′ ( 80 kpc) to the North-East. The color of the cluster red-sequence of Z–H 2.2 ∼ ≃ (Vega) is fully consistent with model predictions (Fioc and Rocca-Volmerange, 1997) of solar metallicity passively evolving galaxies with formation redshift zf 5 (see left panel of Fig. 10.6 and Fig. 7.1). The spectroscopic conﬁrmation of XMMUJ0104.4-0630∼ at a redshift very close to the photometric Z–H red-sequence estimate is hence a successful test for the new method. The BCG of this cluster is within 0.1 mag of the expected color for the passive evolution model as indicated by the green square in the CMD of Fig. 10.6. 196 CHAPTER 10. HIGH-REDSHIFT CLUSTER STUDIES: FIRST RESULTS

ID Name z fX(0.5–2.0 keV) rc LX(0.5–2.0 keV) 10−14 erg s−1cm−2 ′′ 1043 erg s−1 A XMMU J0104.4-0630 0.947 0.005 1.7 0.3 18.8 0.8 6.4 1.3 ± ± ± ± B XMMU J0104.1-0635 0.95 0.05 1.2 0.4 45 4 4.4 1.4 ± ± ± ± Table 10.1: Properties of the two newly discovered X-ray clusters.

Figure 10.6: Color composites and color magnitude diagrams for different target regions of the field. Top left: 2′ 2′ R+Z+H color image of XMMU J0104.4-0630 at z = 0.947 with white X- × ray contours overlaid. Top center: 1.5′ 1.5′ Z+H color composite of the second X-ray selected × cluster XMMU J0104.1-0635 with the same redshift estimate from its Z–H red-sequence . Top right: 1.5′ 1.5′ zoom on one of the X-ray undetected overdensities of bluer galaxies in the center of the field. × Bottom row: Z–H CMDs of the three objects shown in the top panels. Filled dark symbols indicate objects within a 30′′ radius from the center, filled green symbols are from 0.5′–1′, black dots are all objects in the field. Left and center: CMDs of the two X-ray selected galaxy clusters with concordant red-sequence colors. The dashed horizontal lines indicate the predicted model color for passively evolving galaxies at z = 0.95, the dotted red lines define an interval of 0.15 magnitudes around ± the red-sequence model color represented as the red-galaxy population in Fig. 10.7. Spectroscopically confirmed cluster members of XMMU J0104.4-0630 are indicated in the left panel by open symbols. The BCG is marked by the green square, the six additional secure members by green circles, and the eight tentative cluster members by blue open circles. Right: CMD of the galaxy overdensity between the two X-ray clusters with a bluer red-sequence color. The three galaxies there with available spectroscopy are indicated by open circles. The dashed blue line is shown at the color of the apparent locus, the blue dotted lines define the color cuts for the bluer galaxy population in Fig.10.7. 10.2. INDICATIONS FOR LARGE-SCALE STRUCTURE AT Z 1 197 ∼

Figure 10.7: Red and blue galaxy overdensities in the field of XMMUJ0104.4-0630. Left: 14′ × 14′ field-of-view of galaxy overdensities smoothed with a 150 kpc-kernel of the red and blue galaxy populations (as defined in Fig. 10.6) encoded with the corresponding color; X-ray contours are overlaid in white. The color cuts for both populations span a range of 1.5–5 σ relative to the control field in the lower left quadrant (yellow dashed region), which served as the control field. Labels are showing the locations of the two X-ray selected galaxy clusters corresponding to 7-σ peaks in the red population, and optically selected 3-6 sigma peaks (XUGO 1–6) in the blue population. Right: 3.5′ 3.5′ zoom × on the two X-ray clusters.

The cluster center has been identiﬁed as the position of a 1.4 GHz radio source with a ﬂux of 11.9 1.0 mJy in the NVSS survey (Condon et al., 1998), which is likely to be ± 25 −1 attributed to a radio galaxy with a radio power P1.4 GHz 5 10 W Hz . The positional error circle of a few arcseconds radius includes the two∼ brig× htest cluster galaxies. The fairly strong radio emission could be an indication of cooling core activity in the cluster center. In any case, radio emission of this order in high-z clusters is a prime concern for the cluster selection of upcoming SZE-surveys (Lin and Mohr, 2007) and requires further detailed studies (see Sect. 2.3.6 & Sect. 12.1). 198 CHAPTER 10. HIGH-REDSHIFT CLUSTER STUDIES: FIRST RESULTS

Figure 10.8: Galaxy populations of XMMU J0104.4-0630 at z =0.947. Left: 3′ 3′ view on the red × galaxy population smoothed with a 75kpc kernel and displayed in color cuts spanning overdensities of 2–5 σ. Red circles indicate the positions of individual galaxies, the green cross marks the X- ray determined cluster center of the blue contours, and the circle represents the cluster core radius of r 19′′, corresponding to 150 kpc. Right: Same FoV and representation for the blue galaxy c ≃ population.

Detection of the second X-ray cluster XMMU J0104.1-0635

The CMD of the second X-ray selected cluster XMMU J0104.1-0635 reveals a red-sequence color fully consistent with the same redshift as the main cluster A, but with a fainter galaxy population. The overdensity of red galaxies (see Fig. 10.7) selected with a (red) color-cut of 2.23 0.15 (corresponding to the photometric color uncertainty at H 20.2) shows a ± ∼ 7 σ peak (6.5 σ for cluster A) relative to the mean density and standard deviation in the 45 square arcminute control field in the lower left quadrant. We thus assign a photometric red-sequence redshift of z =0.95 0.05 to cluster B. We propose that cluster A and B are ± physically associated with each other, forming a double cluster with a projected separation in the plane of the sky of 3 Mpc. A literature redshift of z =0.932 from the XMM-Newton Serendipitous Survey (Barcons et al., 2002) for a BLLac object 75′′ to the South-West of cluster B supports the idea of a large-scale structure filament along the axis of the two systems (see lower right corner of Fig. 10.5). Currently only two X-ray luminous double clusters at higher redshifts are reported in the literature (e.g. Hashimoto et al., 2005; Nakata et al., 2005). The photometric identification of XMMU J0104.1-0635 confirms the efficacy of the X-ray detection and selection procedure (Chap. 6) even for faint extended sources with large core radii (see Sect. 9.1.3). 10.2. INDICATIONS FOR LARGE-SCALE STRUCTURE AT Z 1 199 ∼

Associated large-scale structure environment The center of the proposed 3 Mpc-scale cluster-cluster-bridge is marked by an additional significant overdensity of slightly bluer galaxies, with very weak X-ray emission well below the detection threshold for extended sources. The color composite and CMD of this X- ray Undetected Galaxy Overdensity (XUGO) are shown in the right panels of Fig. 10.6 revealing a tight red-sequence at a 0.25 mag bluer color. Without additional spectroscopy, there are two plausible scenarios for∼ this object: (i) XUGO 1 is a foreground group or low- mass cluster at z 0.7 based on the observed Z–H color, or (ii) this galaxy overdensity is ∼ part of the LSS at z 0.95 with a systematically bluer galaxy population. This speculative interpretation is supported≃ by three tentative cluster redshifts within 30′′ (see right lower panel of Fig. 10.6) and the suggestive geometric alignment with the two X-ray clusters. In order to investigate the nature of this system, we set a second (blue) adjoint color- cut at 1.8 Z H < 2.08, corresponding to the blue dotted lines in Fig. 10.6. The spatial distribution≤ of− color selected objects was smoothed with a Gaussian kernel of 150 kpc physical scale (19′′), the approximate core radius of cluster A. The resulting overdensity plot of the 14′ 14′ field-of-view is shown in Fig. 10.7, displayed with color cuts ranging from 1.5 to 5 σ×overdensity relative to a control field in the lower left quadrant. The two brightest peaks in the red galaxy population mark the X-ray selected clusters and are well- centered on the diffuse X-ray emission. On the other hand, the optically selected system (XUGO1) corresponds to a 4.5-σ peak of the bluer population. Figure. 10.7 (left panel) shows that this population can be traced to the outskirts of the main cluster A, where blue 5-σ peaks to the NE and SW enclose the cluster center along the axis laid out by the X-ray morphology. The drop to 3.5 σ in the very center of cluster A in conjunction with the rising red overdensity to 6.5 σ (see Fig. 10.7, top right) suggests that this is a direct consequence of the proposed scenario of delayed star formation quenching in lower density environments (e.g. Thomas et al., 2005; Cucciati et al., 2006). Under the burst model assumption, the observed 0.15–0.3 mag Z–H color shift towards the blue is consistent with an age difference of the stellar populations of 1.2–2.2 Gyr. In Figure 10.8, the two different galaxy populations of XMMU J0104.4-0630 are displayed in separate panels showing the red cluster core galaxies in the left panel and the proposed filament population on the right. Systematic blue shifts of the S0 population have been observed in several clusters at lower redshift (e.g. van Dokkum et al., 1998; Abraham et al., 1996) and at z 1.1 (Mei ∼ et al., 2006a). The observed blue population in Fig. 10.7, which includes the spectroscopic cluster members with IDs 3, 4, 6 and two additional tentative members, could hence be interpreted as an evolving S0 population. However, firm conclusions on physically associated filaments and member galaxies of XMMU J0104.4-0630 will only be possible with additional spectroscopic observations of the field. 200 CHAPTER 10. HIGH-REDSHIFT CLUSTER STUDIES: FIRST RESULTS

10.3 Summary and Conclusions

In Sect. 10.1, the Z–H red-sequence method was calibrated using ten spectroscopically • conﬁrmed clusters. It was shown that the (zf = 5, Z = Z⊙)-passive evolution model is fully consistent with all observations and was therefore applied as reference model for the photometric redshift estimation of the full cluster candidate sample.

The observed Z–H red-sequence color evolution over the redshift baseline 0.2

In Sect. 10.2, details of the newly discovered X-ray luminous cluster of galaxies • XMMU J0104.4-0630 at redshift z = 0.947 0.005 were presented. The compact, intermediate mass cluster is found to be in an± evolved state and hosts a strong central radio source. It was shown that the cluster shows a pronounced stratification of galaxy populations. Whereas the spatial distribution of the red-sequence population of early type galaxies coincides well with the X-ray emission, a significantly bluer population dominates beyond 1–2 core radii from the center, suggesting a cluster environment-driven effect of differential galaxy evolution, e.g. delayed star formation quenching in the outskirts.

The Z–H color of the cluster red-sequence was shown to be in good agreement with • the predictions of the reference model and conﬁrmed the applicability of the method for reliable photometric redshift estimates.

A second X-ray selected cluster, XMMU J0104.1-0635, 6.4′ to the South-West could • be photometrically identiﬁed at z 0.95. It was speculated that this object is part ≃ of the large-scale structure environment of the main cluster, which could include additional optically selected galaxy overdensities. Chapter 11

Preliminary Studies and Science Outlook

The available Z–H follow-up observations aimed primarily at the cluster identification and confirmation of the principal strategy in terms of limiting depth, photometric precision, and the choice of filter bands. However, the application of the calibrated Z–H-redshift relation to the full distant cluster candidate sample allows a preliminary study of the luminosity evolution of the brightest cluster galaxies in the redshift range 0.2 < z < 1.5, presented in Sect. 11.1. Section 11.2 provides a brief outlook on upcoming investigations∼ ∼ of the faint end of the cluster red-sequence. The chapter ends with a gallery of selected, newly discovered high-redshift systems awaiting spectroscopic confirmation.

11.1 The Assembly History of Brightest Cluster Galaxies

In this section, the cosmic evolution of the brightest cluster galaxies is investigated. Back- ground information and the latest simulation results on BCGs have been provided in Sect. 2.4.4. The Hubble diagram (see Equ. 3.46 and Sect. 3.3) will be applied as a diagnostic tool for BCG evolution in a ﬁxed concordance model cosmology. Absolute BCG H-band magnitudes out to z 1.5 are derived and compared to predictions of passive evo- ∼ lution models, which have been shown to be very successful for the description of the Z–H color evolution in Sect. 10.1.

11.1.1 BCG sample and selection The preliminary study presented here uses the full Calar Alto sample (see Fig. 9.4) of X-ray selected clusters, which fulﬁlled these selection criteria: (i) a signiﬁcant detection of an optical cluster counterpart and (ii) a reliable red-sequence redshift estimate with a maximum accepted uncertainty of ∆z =0.2. The remaining BCG cluster sample contains a total of 63 systems between z =0.18 and z 1.5. ≃ 201 202 CHAPTER 11. PRELIMINARY STUDIES AND SCIENCE OUTLOOK

Figure 11.1: Binned projected cluster-centric distance histogram of the 63 selected BCGs with respect to the X-ray center position. Top: The BCG locations of the low and intermediate redshift bin with z<0.8 (38 objects) are predominantly coincident with the X-ray marked cluster center. Bottom: At high redshifts of z 0.8 (25 clusters) the distribution is skewed towards slightly larger oﬀsets from ≥ the X-ray center. Extreme outliers in both histograms can be mostly attributed to X-ray center biases due to point source ﬂux contaminations.

The brightest cluster galaxies are selected based on their H-band magnitude, in combination with selection cuts concerning the cluster-centric distance, defined with respect to the X-ray center, and the Z–H color offset from the red-sequence. The increasing foreground contamination of galaxies with magnitudes similar to the BCGs of distant clusters requires tightening identification constraints with increasing cluster-centric distance to ensure a robust BCG selection. In the absence of a spectroscopic identification and photometric redshifts derived from deep multi-band data, the BCG selection based on cluster-centric distance and reasonable Z–H cuts is currently the best applicable method for the available data. The following selection criteria were applied for the BCG identification in the 63 X-ray clusters: 1. 21 central dominant galaxies within 5′′ of the X-ray marked center are accepted as BCGs without color cuts; 2. 37 identifications are based on galaxies with cluster-centric distances of 5′′–30′′ and a color cut of 0.5 mag around the red-sequence Z–H color; ± 3. 5 selected BCGs are located 30′′–60′′ off-center but are within 0.3mag of the red- sequence color. ± 11.1. THE ASSEMBLY HISTORY OF BRIGHTEST CLUSTER GALAXIES 203

These selection criteria are motivated by the first experiences with the spectroscopically confirmed sub-sample discussed in Sect. 10.1.2. However, the robustness of these preliminary selection criteria needs to be investigated and quantified with upcoming follow-up observations of high-redshift candidates. The maximum search radius of 60′′ corresponds to about 500kpc at z > 0.8. For the majority of X-ray selected clusters, the identification of BCGs close to the∼ X-ray center is straightforward and fairly unambiguous. The sample contamination level for non-BCG selections is currently expected to be no more than of order 10%, i.e. sub-sample averages should be statistically robust. Figure 11.1 displays the histogram of binned cluster-centric distances of the identified BCGs for low and intermediate redshift clusters at z< 0.8 (top panel) and for high-redshifts systems at z 0.8 (bottom panel). A slight shift towards higher off-center distances can be observed≥ at high redshift as a first tentative indication that BCGs at large lookback times are not yet in their final configuration with respect to the cluster center.

11.1.2 The BCG H-band Hubble diagram Using the observed apparent total BCG magnitudes and the estimated red-sequence redshifts, we can construct the H-band BCG Hubble diagram out to z 1.5. H-band galactic ≃ extinction corrections AH have been applied to all BCG magnitudes using the galactic reddening maps1 of Schlegel, Finkbeiner & Davis (1998). The final BCG H-band Hubble diagram, i.e. the (mH(BCG) AH) versus log z relation, is shown in the top panel of Fig. 11.2. Black dashed lines indicate− the model expectations for solar metallicity, passive evolution burst models with formation epoch zf =5 for m*, m* 1, m* 2, and m* 3 galaxies (top to bottom). Red arrows mark objects with strong signs− of object− blending− of the SExtractor photometry, i.e. the determined magnitudes are lower limits (implying upper limits of the luminosity). Photometric errors are typically small, except for blended sources, since all BCGs are at least two magnitudes brighter than the limiting H-band image depth. Errors are hence dominated by the redshift uncertainties of the Z–H red-sequence method. The lower panel of Fig. 11.2 shows the same plot with a linear redshift axis, which provides a better illustration of the effects of redshift errors, in particular at low-z where the apparent magnitudes faint rapidly with increasing distance. Instead of assuming standard candle properties of the BCGs and applying the Hub- ble diagram as a cosmological test for parameter constraints (e.g. Collins & Mann, 1998; Aragón-Salamanca et al., 1998), we assume a fixed concordance cosmology and use the Hubble diagram as a tool for BCG evolution studies. Equation 3.46 reveals that variations of the Hubble constant H0 introduce a normalization offset, whereas different values for Ωm, ΩΛ, and w change the shape of the Hubble relation at high redshifts. However, parameter changes within the concordance model uncertainties introduce variations in the luminosity distance d at the high-redshift end z 1.5 of less than 10%. lum ∼

1A galactic extinction calculator is available at http://nedwww.ipac.caltech.edu/forms/ calculator.html. 204 CHAPTER 11. PRELIMINARY STUDIES AND SCIENCE OUTLOOK

Figure 11.2: Top: H-band Hubble diagram for the BCGs of 63 X-ray selected clusters between redshifts of z =0.18 and z 1.5 . Dashed lines indicate solar metallicity, passive evolution burst model ≃ expectations for a formation epoch of z =5 and galaxy luminosities of m*, m* 1, m* 2, and m* 3 f − − − (top to bottom). BCGs with photometry aﬀected by blending are marked by red arrows. Bottom: Same plot with a linear redshift axis. 11.1. THE ASSEMBLY HISTORY OF BRIGHTEST CLUSTER GALAXIES 205

Figure 11.3: Redshift evolution diagram of the absolute H-band BCG magnitudes with estimated 1 σ errors. The observations are fully consistent with a non-evolving absolute magnitude of M = 26.3 mag (red dashed line), with an rms scatter of 0.6 mag (red dotted lines), which is mostly H − introduced by the photometric redshift uncertainties. Black dashed lines indicate the expected evolution of passive galaxies, which are expected to dim by about one magnitude from redshift z 1.5 to ≃ the present epoch. Purely passive BCG luminosity evolution can be excluded at the 2.6 σ level.

Figure 11.4: K-correction terms as a function of redshift for a zf = 5 passive evolution BCG model for diﬀerent bands. The applied H-band correction term (red line) is very small at all relevant redshifts. 206 CHAPTER 11. PRELIMINARY STUDIES AND SCIENCE OUTLOOK

BCGs as standard candles As a ﬁrst application, the absolute H-band magnitudes for the BCGs are computed following Equ. 3.46

M (BCG) = m (BCG) A K (z) 25 5log d [h−1 Mpc] . (11.1) H H − H − H − − lum 70 2 The K-correction terms KH to transform observed mH magnitudes at redshift z into rest-frame H-magnitudes have been calculated for the standard zf = 5, solar metallicity passive evolution model and are displayed by the red line in Fig.11.4. Note that the H- band observations exhibit the smallest absolute K-correction terms of all bands, implying that the observed apparent H-band magnitudes are very close to the associated rest-frame values. The BCG redshift uncertainties, as the dominant error source for all non- spectroscopically confirmed clusters, enter the error budget of the absolute magnitudes via dMH = dmH + 5[(d log dlum)/dz] dz. In particular at z < 0.4, where the luminosity distance is a steep function of z (see lower panel of Fig. 11.2),∼ the photometric redshift errors influence dMH in a magnified way. The absolute BCG H-band magnitude diagram of Fig. 11.3, with a full accounting of the error budget included, reveals an astonishing result: the brightest cluster galaxies of X-ray selected clusters are fully consistent with being true standard candles with constant absolute rest-frame H-band luminosity of MH 26.3 mag out to z 1.5. ≃− ≃ This result is statistically very robust. The median and mean values of (i) the full sample of 63 BCGs, (ii) the three lowest redshift clusters (spectroscopically confirmed), and (iii) the five highest redshift objects all agree within 0.06 mag. The scatter of the full ± sample around the mean value MH 26.3 mag is σM 0.6 mag, which is for the largest part to be attributed to the redshift≃− uncertainties. A≃ linear fit to the full BCG sample yields the formal functional form MH = ( 26.23 0.20) + ( 0.10 0.22) z, also fully consistent with a non-evolving absolute H magnitude.− ± − ± · The full BCG sample contains four outliers, all of which have associated literature redshifts, i.e. their estimated MH errors are small. The BCGs of the two clusters A 383 with MH = 26.65 0.1, and RXJ0018+1617 with MH = 27.2 0.1 (see Sect. 10.1 for references) are− likely± truly super-luminous systems compared− to the± sample average. The remaining two outliers are clusters LCDS 0504 with spectroscopic redshift z =0.794 and LDCS 0505 with a photometric literature redshift of z 0.48 from the Las Campanas Distant Cluster Survey (Gonzalez et al., 2001). The BCG≃ of LCDS0504 with M = 27.45 H − is flagged as a photometric blend (see panel 11 of Fig. 11.7). In the case of LDCS0505, with an apparently underluminous BCG of MH = 25.1, the Z–H red-sequence color suggests that the published photometric redshift of z− 0.48 is underestimated. The latter two ≃ outlier objects can thus be attributed to systematic effects rather than significant intrinsic deviations from the mean absolute BCG magnitude.

2Models computed by Daniele Pierini. 11.1. THE ASSEMBLY HISTORY OF BRIGHTEST CLUSTER GALAXIES 207

The fact that BCGs are good standard candles has long been established for redshifts of z<1 (see Sect. 2.4.4). The new preliminary result from the sample of this work with the increased redshift baseline out to z 1.5 is the conclusion that the observations are not consistent with pure passive evolution≃ . The dashed black lines in Fig. 11.3 indicate the expected fading of the absolute H-band luminosity for passively evolving galaxies by one magnitude from z 1.5 to z =0, corresponding to an average slope of dMmodel/dz 0.67. Comparing this result≃ to the constraints obtained from the linear ﬁt to the observed≃− data of dM /dz 0.10 0.22, excludes a purely passive BCG luminosity evolution at the H ≃ − ± 2.6 σ level.

BCG luminosity evolution As the next step, the BCG luminosity evolution relative to the passively evolving model is investigated. The apparent magnitudes of the standard zf = 5, solar metallicity model for ∗ an Lmod galaxy, mmodel, are subtracted from the extinction corrected observed magnitudes 3 mH(BCG) AH. The results are shown in Fig. 11.5, with the same meaning of the symbols − ∗ as before. In this representation, passively evolving Lmod galaxies would be located along the null line (top horizontal dashed line). The bulk of the BCG population is observed in the diagram region in between galaxies one magnitude (center dashed line) and two magnitudes ∗ (lower dashed line) brighter than Lmod. The errors in the model-subtracted magnitudes originate from the redshift uncertainties of the BCGs, m (z ∆z), in analogy with model ± Fig. 11.3. Since Fig. 11.5 displays the magnitude residuals after accounting for passive evolution, more negative data points at low redshifts can be interpreted in terms of an increased luminosity, in excess of passive evolution, compared to the high-redshift BCGs. As a first approach for modelling the residual BCG luminosity evolution, we can assume a linear relationship with redshift of the form mH(BCG) mmodel =A0+B z, with fit parameters A0 for the local magnitude difference, and B for the− slope. A linear fit· to the full BCG sample over all redshifts yields the parameter constraints A = 2.0 0.2 and B =0.60 0.22, 0 − ± ± shown by the green line in the top panel of Fig. 11.6. Black symbols with error bars represent the observed BCG data, with one outlier to either side not displayed in the zoomed representation of the central residual magnitude interval of interest. Blue dots symbolize the magnitude and redshift medians of 13 independent bins of five BCGs in order of decreasing redshift (three objects for the lowest z bin). The medians are less affected by outliers compared to averaging, e.g. due to blends in the photometry, and should represent a robust smoothed trend of the observed data as visual guide. A linear fit to the median smoothed data results in consistent parameters but with decreased uncertainties, ruling out a flat model with passive evolution at the 3.7 σ level, whereas the raw data fit yields a 2.7 σ confidence level.

3 The galactic extinction correction term AH is omitted in the axis label for better clarity, mH(BCG) implicitly denotes extinction corrected magnitudes. 208 CHAPTER 11. PRELIMINARY STUDIES AND SCIENCE OUTLOOK

Figure 11.5: Observed apparent magnitude evolution relative to expectations of passive evolution ∗ ∗ models. The expected apparent model magnitudes mmod(z) for an Lmod galaxy have been subtracted, implying that the two dashed lower lines represent galaxies one (center line) and two (bottom line) magnitudes brighter than M*.

A linear relationship would imply that 2/3 of the luminosity evolution happened at redshifts below unity, at variance with observations (e.g. Burke et al., 2000). The apparent discrepancy can be resolved by splitting the BCG sample into a low–intermediate redshift bin at z 0.9, approximately corresponding to the range accessible by former BCG analyses, and≤ a high-redshift bin with 18 objects with z> 0.9. The linear fits to the two sub-samples yield the parameter constraints A = 1.7 0.4, B =0.0 0.5 for the low 0 − ± ± redshift bin, and A0 = 2.2 0.5, B = 0.8 0.4 for the independent high-z sub-sample, displayed by the red lines− in± the top panel− ± of Fig. 11.6. The observed low–intermediate redshift BCGs at z 0.9 are hence consistent with a passive evolution model, although with increased errors,≤ in agreement with previous findings. At z> 0.9, a significant jump in the residual BCG magnitudes is apparent. Note that the non-passive evolution signal is present for the three rightmost blue points, representing the median values of the 15 highest redshift BCGs. The observed effect is hence robust against the removal of individual high-z objects from the sample. 11.1. THE ASSEMBLY HISTORY OF BRIGHTEST CLUSTER GALAXIES 209

Figure 11.6: BCG luminosity evolution models compared to observations. Top: Linear fits to the data over the full sample (green line), and separate linear models for z 0.9 and z>0.9 sub-samples ≤ (red lines). Black background symbols show the observed data, blue circles represent the medians of independent bins of five objects. Bottom: Maximum likelihood model fit for a step function with a softened transition zone given by the red solid line, dashed lines display 1 σ variations of the fit parameters. The observations are consistent with a mass assembly phase of the BCGs centered at z 0.95. Since their colors evolve as for passively evolving stellar populations, the increase in ∼ luminosity is driven by an increase in stellar mass. 210 CHAPTER 11. PRELIMINARY STUDIES AND SCIENCE OUTLOOK

The observed discontinuity for the split sample suggests the application of a step function model with a gradual transition zone, in analogy to a Fermi function. The functional form can then be parameterized as ∆m mH(BCG) mmodel =(A0 + ∆m) z−za , (11.2) − − 1 + exp( b ) where A0 is the local magnitude residual, ∆m the step height, za the central BCG ‘assembly’ redshift, and b a softening parameter that determines the width of the transition zone. Since the data is not sufficient to provide meaningful constraints for a functional fit with four free parameters, the end points are fixed at A0 = 1.91 (at z 0.2) and ∆m =0.86, consistent with the results of a constant absolute luminosi− ty in Fig.≃ 11.3 and the linear fit in Fig. 11.6 (green line in top panel). A maximum likelihood fit to the data4 with two remaining free parameters yields the constraints z =0.96 0.05 for the central ‘assembly’ redshift and b =0.19 0.07 for the a ± ± transition zone width. The solid red line in the bottom panel of Fig. 11.6 displays the best fit model, with the effects of 1 σ parameter variations shown by the dashed red lines. The reduced χ2 value of 1.78 for this softened step function model is improved compared to the linear fit (green fit) with a χ2 of 1.89. The luminosity evolution fit shown in the lower panel of Fig. 11.5 is currently the best preliminary model for the available BCG data, consistent with only small deviation from passive evolution at low redshifts as reported by former studies.

Biases and systematics We will now address the question whether the observed deviations from passive BCG luminosity evolution models could be an artefact introduced by systematic eﬀects and selection biases. The main areas of concern are brieﬂy discussed in the following:

Non-Representative Cluster Sample: Using a heterogeneous cluster sample with varying object properties as a function of redshift can introduce biases that mimic cosmic evolution. In fact, Burke, Collins & Mann (2000) showed that the initial claim of a non-passive BCG luminosity evolution of Aragon-Salamanca, Baugh & Kauffmann (1998) could be attributed to the bias of mixing an optically selected high-z sample and an X-ray selected low-redshift sample. Owing to the observed weak correlation of ∼1/4 cluster mass and BCG luminosity LBCG M200 (see Sect. 2.4.4), the target selection from high-redshift clusters with systematically∝ lower mass, and hence lower X-ray luminosity, introduces the selection artefact of intrinsically fainter distant BCGs. The cluster sample presented here is X-ray selected at all redshifts, introducing a general selection bias with the opposite sign, i.e. an effective ‘anti-bias’. At a given approximate flux limit, the minimum cluster luminosity is increasing with redshift

4The magnitude errors of the four outliers identiﬁed in the absolute magnitude diagram of Fig. 11.3 have been ’softened’ to 1 σ deviations to prevent the likelihood ﬁt being dominated by individual objects (with systematics). 11.1. THE ASSEMBLY HISTORY OF BRIGHTEST CLUSTER GALAXIES 211

> 2 according to LX 4πfX,lim dlum, implying an increasing minimum cluster mass for the 58 serendipitously∼ detected objects. The five additional XMM target clusters (see Sect. 9.2) at low and intermediate redshifts are high X-ray luminosity, massive clusters, of which three were identified as BCG outliers on the bright end. The X-ray selection bias of finding predominantly low mass clusters and groups at lower redshifts (e.g. Pacaud et al., 2007) contributes to the observed increased scatter at low-z in Fig. 11.3. The detected non-passive BCG luminosity evolution of ∆m 0.9 mag is hence to be interpreted as a lower limit, with a larger effect to be expected≃ using a representative cluster sample. In other words, we find distant BCGs to be fainter than expected from passive evolution, despite the fact that the high-redshift objects have an X-ray selection bias towards intrinsically brighter absolute magnitudes, i.e. we observe a positive slope in Fig. 11.6 but would expect a negative correlation. The X-ray selection bias effectively increases the confidence in the detection of non-passive BCG evolution, but has to be taken into account for the comparison with unbiased predictions from simulations.

BCG Selection: The BCG selection has been discussed in Sect. 11.1.1. While individual misclassifications based on the two-band data are possible, the sample statistics should be robust. A significant bias is only expected if a large fraction of distant BCGs is either located at cluster-centric distances of >500 kpc, or the BCG color is more than 0.5 mag bluer than the red-sequence; both scenarios seem unlikely based on known cluster populations. Since the BCG identification is more difficult at high redshift, the current approach needs to be justified once additional data is available leading to possibly refined selection criteria.

Cluster Redshift Estimates: To be consistent with passive evolution models, the redshifts of the high-z clusters would have to be signiﬁcantly underestimated (see lower panel of Fig. 11.2). This would be nice from the distant X-ray cluster discovery point- of-view, but is very unlikely given the high redshift-sensitivity of the Z–H technique at z > 0.9 (see Sect. 7.1.3). Using the ‘red-envelope’ in the color-magnitude diagram as distance∼ indicator for high-z clusters (see Sect. 10.1.2) could possibly lead to a slight overestimation of the redshift for some distant clusters, which would again lead to an anti-bias in the sense that the distant BCGs are intrinsically even fainter than measured. Note that the spectroscopically conﬁrmed cluster XMMXCS J2215.9 1738 at z =1.45 exhibits the faintest BCG of the whole sample. −

Cosmological Model: As discussed, a modified Hubble parameter H0 only changes the normalization of the Hubble diagram, but not its shape (see Equ. 3.46). Varying Ωm within reasonable limits of 0.2 < Ωm < 0.4 changes the luminosity distance of a flat ∼ ∼ cosmology dlum by less than 10% at z 1.5, which would result in apparent magnitude differences at the highest redshifts of∼ ∆m< 0.2 mag (see Equ. 7.4), i.e. less than 1/4 of the observed magnitude offset. ∼ 212 CHAPTER 11. PRELIMINARY STUDIES AND SCIENCE OUTLOOK

Cosmological Surface Brightness Dimming: The severe cosmological surface brightness dimming (Equ. 3.22) can introduce signiﬁcant selection biases for very distant (z > 2) galaxy studies close to the detection limit (Lanzetta et al., 2002). How- ever,∼ all galaxies of the BCG sample are at least two magnitudes above the H-band completeness limit, implying that surface dimming should not play a signiﬁcant role.

Biased Photometry: The H-band BCG data is taken from a homogeneous data set, observed with the same instrument and reduced in a standardized way. Great care has been taken to optimize the NIR sky subtraction to avoid background biases, in particular for faint, low surface brightness objects (see Sect. 7.3). The SExtractor photometry software is well tested and robust at the BCG magnitudes well above the detection threshold. SExtractor systematics due to source blending effects have been discussed and are marked in diagrams by arrows to indicate lower limits. Without source blending, the affected BCGs in the range 0.5 > z > 1.1 could be intrinsically slighter fainter, with minor effects on the overall form∼ of∼ the luminosity evolution.

Passive Evolution Models: BCGs typically exhibit red, old stellar populations, and are in most cases the top ranked galaxies on the red-sequence (see Sect. 10.3). The assumption that the stellar populations of BCGs evolve passively seems therefore to be well justified. In Sect. 10.1, we could show that the red-sequence Z–H color evolution out to the highest accessible redshifts is fully consistent with passive evolution models with an early formation epoch. In addition, the X-ray selection process disfa- vors clusters with central (strongly non-passive) AGN activity with their associated point-like X-ray flux contribution. For the BCGs in our X-ray selected, red-sequence confirmed cluster sample passive evolution is a good description for the colors of their stellar populations. This means that recent star formation episodes or moderately strong active nuclei are expected to play a minor role.

For the comparison to passive evolution expectations, the standard zf = 5, solar metallicity model is applied, which was shown to be close to the best fit parameters of Fig. 10.2. Variations of the formation age result in a slightly different mH evolution over the observed redshift range from z 1.5 to z 0.2. At a fixed local apparent ∼ ∼ H-band magnitude, a zf = 3 passive galaxy is expected to be about 0.2 mag brighter at z 1.5 compared to the standard model, whereas formation redshift zf = 10 objects∼ are 0.2 mag fainter. This implies that subtracting passive evolution models with zf < 5, as derived in Sect. 10.1, results in an increased magnitude offset at high redshifts, since the BCG magnitudes are compared to brighter model galaxies. The observed non-passive BCG luminosity evolution will hence be enhanced, if the formation redshift is lowered.

Although a number of possible biases and selection effects could be identified, all individual sources of error can at most, under reasonable assumptions, account for 1/4 of the observed non-passive component shown in Fig. 11.6. The main aim of this preliminary study at this point is to provide conclusive direct observational indications, rather than final exact 11.1. THE ASSEMBLY HISTORY OF BRIGHTEST CLUSTER GALAXIES 213

values, that BCGs are indeed increasing their mass as expected from hierarchical galaxy formation models, whereas their stellar populations evolve passively. For an accurate quantitative measurement of the BCGs evolution, all of the discussed possible biases and systematic eﬀects need to be carefully taken into account for detailed follow-up studies. The preliminary total observed non-passive luminosity evolution component of ∆m 0.9 ≃ is likely a lower limit, due to the discussed anti-bias of the X-ray cluster selection and the use of a zf =5 model for comparison.

11.1.3 Observing the BCG assembly epoch Under the assumption that BCGs are assembled from progenitors with passively evolving stellar populations, i.e. associating the luminosity gain to additional mass, the detected evolution should be accompanied by an increased merging activity at z 1. The following sub-section will take a closer look at this scenario. ∼

Merging activity Before addressing the question of BCG merging events at high redshifts, we need to approximate the occurrence rate of chance projected alignments of galaxies in clusters. Assuming that the BCG is located close to the center and that the galaxy density in the inner core region is approximately constant, then the probability that a second detected cluster galaxy is within the projected radius rproj of the BCG can be estimated as

2 Vproj rproj pproj Ngal 1.5 Ngal < 0.1 , (11.3) ≃ · Vtot ≃ rtot ∼ where V 2πr2 r is the projected cluster volume within r along the cluster proj ≃ proj tot proj diameter 2 rtot, Vtot is the approximate total volume of the inner cluster region, and Ngal is the number of detected cluster galaxies. For distant objects, observed at seeing limited ′′ ′′ 1 resolution, we can set the radius for apparent projected encounters to rproj 3 , i.e. ∼2–3 times the seeing, the cluster radius for the denser core to r 0.5 Mpc ≃60′′, and tot ≃ ≃ the number of detected galaxies at the given image depth to Ngal 20, yielding a projected encounter chance of about 8%. This value is a ﬁrst crude estimate≃ of the occurrence rate of close galaxy pairs due to chance alignments in cluster cores at the observed image depth, but does not include additional foreground and background objects yet. For analytic approximations of chance alignments, values for the considered outer cluster radius rtot and the total number of detected objects Ngal have to be assumed, which are not accurately known. As an alternative method without strong assumptions, the observed occurrence rate of close pairs for non-BCGs galaxies in the cluster core can be determined and compared to the BCG values. In 1′ 1′ environments surrounding the BCGs at 0.9 < z < 1.2, about 15% of the non-BCG galaxies× are found as close pairs with separations of∼<∼3′′. Since this method also accounts for real mergers, the value can be considered as a∼ conservative reference point. In summary, an occurrence rate for close 214 CHAPTER 11. PRELIMINARY STUDIES AND SCIENCE OUTLOOK

galaxy pairs of 10–15% can be accounted for by chance alignments at the image depth and resolution limit of the available data. Figures 11.7 & 11.8 display a gallery of 24 selected BCGs (marked by green arrows). The ordering of panels 1–24 is a speculative attempt to place the observed BCG appearance along an evolutionary sequence, i.e. to provide ‘typical BCG snap-shots’ over the last 9 Gyr of cosmic time. At the earliest observationally accessible epochs (e.g. panels 1–3), the BCGs appear to be ordinary galaxies very similar to their neighbors. At redshifts of z < 1.3, almost 40% of all BCGs exhibit indications for possible ongoing merging events (e.g. panels∼ 4–15), i.e. 2–3 times the value that can be accounted for by projection eﬀects. While a few of these observed close galaxy encounters are expected to be chance alignments (e.g. panel 9), the majority is likely to be attributed to real BCGs merging events, in particular for the cases with more than two distinguishable objects (e.g. panels, 5, 6, 10-15). The apparent merger fraction remains high down to redshifts of z 0.5, but the physical mechanism responsible for the close encounters appears to change,∼ as radially infalling subgroups reach the center (e.g. panels 16–19). In the ﬁnal phase of the apparent BCG evolution, the central dominant giant cD galaxies appear on the cosmic stage at redshifts z < 0.6 (e.g. panels 20–24 in Fig. 11.8). ∼

The four phases of BCG assembly Since the BCGs constitute the most homogeneous class of galaxies in the local Universe (see Sect. 2.4.4), it can be expected that their formation history is very similar. Based on the ﬁndings discussed in this section, the following speculative, observationally motivated formation scenario for BCGs is proposed, complemented by results of numerical simulations (see Sect. 2.4).

Phase A – Normal Galaxies at High Redshift: At the presently observationally highest accessible redshifts of z>1.3, BCGs (5 in the current sample, e.g. panels 1–3 in Fig. 11.7) seem to be ordinary∼ galaxies with H-band magnitudes of approximately m* 1, consistent with a top ranked galaxy in the cluster luminosity function. Together with− other massive cluster galaxies, BCGs migrate towards the bottom of the cluster potential well via the dynamic friction mechanism (see Sect. 2.4.1).

Phase B – Main Merging and Mass Assembly Epoch: During the cosmic epoch 1.3 >z>0.6–0.7, corresponding to lookback times between about 9 and 6 Gyr, the main∼ BCG∼ luminosity and mass assembly takes place through dynamic friction driven dry major merging events close to the cluster center. During this phase, the BCG roughly doubles in mass and luminosity (e.g. panels 4–12) and becomes the dominant cluster galaxy.

Phase C – Radial Infall of Groups and High Speed Encounters: The central BCG is further fed by radially infalling groups of galaxies with small impact parameters (e.g. panels 13–19) in the redshift range 0.8>z>0.3–0.4. In contrast to the friction induced, slow encounters of phase B, the high∼ impact∼ velocities 11.1. THE ASSEMBLY HISTORY OF BRIGHTEST CLUSTER GALAXIES 215

of the order of the cluster velocity dispersion prohibit direct merging (see Sect. 2.4.1). It seems likely that radially impacting galaxies are (partially) disrupted and the debris dispersed throughout the cluster core, from where it can settle back onto the BCG to eventually form the extended cD envelope (e.g. panels 17–22). The mass growth rate during this phase is slowed down, but the characteristic appearance of local cD galaxies is likely shaped in this evolutionary phase.

Phase D – Settlement and Equilibrium Configuration: Once the stellar debris has settled back onto the central BCG and groups have coalesced with the main cluster, the BCG can assume its final equilibrium configuration as a giant dominant cD galaxy at redshifts z<0.6 (e.g. panels 23–24). ∼ This speculative BCG formation picture provides a testable starting point for further, more detailed studies. In particular, the physical mechanism for the formation of the characteristic cD halo requires additional in-depth investigations. The current BCG sample is based on follow-up observations of about one quarter of the XDCP distant cluster sample, implying that significant progress is possible over the next few years using extended BCG samples in conjunction with spectroscopic and high-resolution studies.

Discussion and comparison with simulations The NIR rest-frame H-band luminosity of galaxies is dominated by old stars and is therefore a good proxy for the total stellar mass. Under the assumption of passively evolving BCG stellar populations, we can obtain a first order photometric stellar mass estimate by relating ∗ the observed averaged apparent H-band magnitude to the model predictions for an Lmod 11 galaxy with a total stellar mass of Mtot(L )=2.84 10 M⊙ . For the best fit model of Fig. 11.6 (bottom panel), the averaged BCG∗ magnitudes× are m* 1.05 at z 1.5, m* 1.4 − ∼ − at z 1.0, m* 1.7 at z 0.5, and m* 1.9 at z 0.2. These magnitudes yield the ∼ − ∼ − BCG ∼ 11 estimated stellar mass evolution for BCGs of Mtot 7.5 10 M⊙ at z 1.5, increasing to BCG 12 BCG ≃12 × ∼ Mtot 1.0 10 M⊙ by z 1.0 and Mtot 1.4 10 M⊙ at z 0.5, with a final mass of BCG ≃ × 12 ∼ ≃ × ∼ Mtot 1.7 10 M⊙ at z 0.2. The observed luminosity evolution from redshift z 1.5 to z 0≃.2 can× therefore be translated∼ into a stellar mass increase by a factor 2.2≃over this≃ cosmic epoch. As discussed, this is to be interpreted as a lower limit for the stellar mass evolution due the anti-bias of the X-ray selection. We can now compare our preliminary observational results to the BCG evolution predicted by semi-analytic hierarchical formations models summarized in Sect. 2.4.4. The left panel of Fig. 2.7 shows that the studied model BCG experiences a major mass assembly epoch around z 1, roughly coincident with the observationally identified epoch of high ∼ BCG merging activity (phase B). For the averaged BCG population in the right panel of Fig. 2.7, the simulations predict a more gradual stellar mass increase by a factor of three from z 1 to the present epoch, or a factor of four for the observationally probed redshift interval≃ 1.5 > z > 0.2. The estimated lower limits for the BCG mass evolution are thus within a factor∼ of∼ two in agreement with the predictions. A similar offset is found when the crudely determined total stellar mass estimates are compared to the median BCG mass predictions of Fig. 2.9. 216 CHAPTER 11. PRELIMINARY STUDIES AND SCIENCE OUTLOOK

Figure 11.7: High-redshift BCG gallery. The panel number, the apparent BCG evolutionary phase indicated by the capital letter (or an intermediate stage), and estimated redshifts are given in the upper left. BCGs are marked by green arrows, X-ray contours are given in blue. All images show H-band or combined Z+H data with identical angular scales. 11.1. THE ASSEMBLY HISTORY OF BRIGHTEST CLUSTER GALAXIES 217

Figure 11.8: Low-redshift BCG gallery. Labels and data are the same as in Fig. 11.7. 218 CHAPTER 11. PRELIMINARY STUDIES AND SCIENCE OUTLOOK

The predicted BCG K-band evolution in Fig. 2.8 can be directly compared to the observed absolute H-band magnitudes in Fig. 11.3 and the model subtracted magnitude evolution in Fig. 11.6. If the X-ray cluster selection anti-bias is taken into account by restricting the simulated high-redshift sample to the brightest BCGs, i.e. the lower part of the sample bins in Fig. 2.8, the agreement with the simulations is remarkable and well within the uncertainties. The preliminary BCG sample study presented here hence manifests a signiﬁcant step forward in the convergence process between observed properties of the most massive galaxies and the advancing predictions of the standard hierarchical galaxy formation paradigm.

11.2 Cluster Red-Sequences Studies

As discussed in Sect. 2.4.3, the effects of the ‘downsizing’ scenario have direct consequences on the red-sequence galaxy population and would result in a ‘truncation’ of the CMR beyond a redshift dependent observed magnitude. The most distant cluster for which a deficit in the faint red galaxy population has been reported is RDCSJ1252.9-2927 and its surrounding LSS at z =1.24, recently investigated in detail by Tanaka et al. (2007). While the main cluster exhibits a clear CMR down to magnitudes of Ks(AB) 23 mag, ∼ the red-sequence of the identified associated clumps in the LSS seem to be truncated at Ks(AB) < 22 mag, consistent with the first reports of this effect at similar redshifts (Kajisawa et∼ al., 2000; Nakata et al., 2001). The observations support the picture of an environment dependent downsizing evolution, where the star formation activity ceases first for the most massive galaxies in clusters and a delayed CMR build-up in lower density environments. At redshift z < 0.8, De Lucia et al. (2007) find a deficit of faint red cluster galaxies in the ∼ luminosity range 0.4>Lfaint/L >0.1 relative to the fraction of Llum > 0.4 L objects. This ∼ ∗∼ ∗ luminous-to-faint galaxy number ratio, Nlum/Nfaint, exhibits a decreasing trend towards lower redshift, implying that the faint end of the CMR becomes increasingly populated with increasing cosmic time. A similar evolution in the build-up of the faint red-sequence galaxy population is observed by Stott et al. (2007) at z <0.5. The current limited depth of the available Z and H-band∼ follow-up data does not yet allow a quantitative assessment of the galaxy population at the faint end of the red- sequence. The detailed analysis of the red-sequence properties at z > 1 requires deep, high- quality multi-band data for precision photometry and the selection of∼ cluster members using photometric redshift techniques, currently available for only about half a dozen clusters listed in Tab. 9.1. The detailed investigation of the ‘downsizing’ scenario and the related red-sequence truncation in galaxy clusters at high redshift will be a main XDCP science focus in the near future. Using guaranteed observing time at the 2.2 m telescope at the La Silla Ob- servatory, we will use the multi-channel GROND5 camera to obtain deep optical and NIR photometric coverage for a sample of z > 0.9 cluster candidates currently scheduled for ∼ 5 More details can be found at http://www.mpe.mpg.de/∼jcg/GROND. 11.3. GALLERY OF HIGH-REDSHIFT CLUSTERS 219 spectroscopic confirmation. GROND provides the unique possibility of performing simultaneous observations in four optical filters and the J, H, and Ks NIR bands. This data will be used to determine accurate photometric redshifts and construct CMDs to completeness limits that are 1–2 magnitudes fainter than the currently available data. In order to fully understand the build-up of the red-sequence and its evolution, the CMRs of distant clusters have to be studied as a function of redshift and cluster mass. Two of the important questions that can be addressed with the upcoming XDCP study are (i) whether the truncation of cluster red-sequences at high redshift is a universal phenomenon and (ii) how the cluster mass influences ‘downsizing’.

11.3 Gallery of High-Redshift Clusters

Before concluding, color composite images of a selection of newly discovered high-redshift clusters are shown in Fig. 11.9 for the redshift range 0.8 < z < 1.05 and in Fig.11.10 for candidate systems at z > 1.1. All 2.′5 2′ color images have∼ ∼ X-ray contours overlaid in yellow and are based on Z+H∼ data obtained× during the two Calar Alto follow-up campaigns. The displayed systems are not yet spectroscopically conﬁrmed, but several of the cluster candidates are already scheduled for FORS 2 spectroscopy. The redshift conﬁrmation is is of particular importance (and a challenge, see Chap. 8) for candidates at the highest accessible redshifts of z > 1.4. Three Calar Alto selected candidate systems belong to the latter category, based on∼ typically 4–6 very red objects close to the X-ray center, two of them are displayed in panels 11 & 12 in Fig. 11.10.

11.4 Summary and Conclusions

In Sect. 11.1, the photometric Z–H red-sequence redshift estimates were used for a • preliminary study of the brightest cluster galaxy evolution by constructing the H- band Hubble diagram for BCGs out to z 1.5. ≃ It was shown, that the observed X-ray selected cluster BCGs are consistent with being • ‘standard candles’ of absolute H-band magnitude M 26.3mag out to z 1.5. H ≃− ≃ The increased redshift baseline allowed to exclude that the luminosity of BCGs • evolves passively at the 3 σ confidence level. Although a number of possible selection biases were identified, which need to be carefully quantified in future studies, none of these effects could fully account for the observed non-passive evolution component.

The comparison to the expected apparent magnitude evolution of a single model • galaxy with passively evolving stellar populations revealed the tentative result that BCGs have at least doubled their mass between redshifts of z 1.5 and z 0.2, ≃ ≃ whereas their stellar populations evolve passively. 220 CHAPTER 11. PRELIMINARY STUDIES AND SCIENCE OUTLOOK

It was speculated that the tentatively identiﬁed active BCG assembly phase around • z 1 is accompanied by an increased observed merger rate at the cluster center, ∼ followed by an epoch where radially infalling groups of galaxies seem to be important for the ﬁnal stages of the BCG evolution.

The preliminary results suggest that the well-established old age of the stellar popula- • tions of BCGs and the expected late mass assembly can be observationally reconciled. Predictions for the BCG evolution from the latest simulations and semi-analytic models are in qualitative agreement with our tentative ﬁndings.

Upcoming deep multi-band photometric studies of XDCP clusters at high redshift • will allow a systematic investigation of the phenomenon of a red-sequence truncation and the related ‘downsizing’ scenario.

A gallery of newly discovered X-ray luminous high-redshift galaxy cluster candidates • was presented. These and other systems are currently awaiting their spectroscopic redshift conﬁrmation.

The reﬁnement of the early results of Chap. 10 and the preliminary BCG evolu- • tion study presented in Sect. 11.1 will provide observational contributions towards an emerging consistent picture of galaxy evolution in clusters. 11.4. SUMMARY AND CONCLUSIONS 221

Figure 11.9: High-redshift galaxy cluster gallery A. 2.′5 2′ Z+H color composites of a selection × of newly discovered high-redshift clusters with photometric redshift estimates of 0.8 < z < 1.05 are shown; X-ray contours are overlaid in yellow. ∼ ∼ 222 CHAPTER 11. PRELIMINARY STUDIES AND SCIENCE OUTLOOK

Figure 11.10: High-redshift galaxy cluster gallery B, displaying a selection of the new sample of candidate clusters with photometric redshift estimates of z >1.1. ∼ Chapter 12

The Future of Galaxy Cluster Surveys

Before closing this work, a brief outlook of some of the expected major developments in distant galaxy cluster studies over the next decade(s) is presented.

12.1 SZE Surveys

A new generation of galaxy cluster survey instruments based on the Sunyaev-Zeldovich effect (SZE) is currently resuming routine operations. 35 years after the theoretical prediction (Sunyaev and Zeldovich, 1972), the SZE (see Sect. 2.3.6) is opening up a new observational survey window to cluster studies. The nearly redshift independent sensitivity to massive clusters and the complementarity with respect to X-ray measurements have raised high hopes for its impact on cosmological studies. The scientific focus of the large SZE surveys is the determination of the cluster number density evolution out to high redshifts as a means to constrain the properties of Dark Energy (Sect. 3.3). The cosmological applications are hence similar to the XDCP goals with the main differences that the SZE selected cluster samples (i) follow a shallower, less peaked redshift distribution compared to the X-ray selection (see right panel of Fig. 6.17), and (ii) have a contiguous, wider sky coverage allowing a clustering analysis. The South Pole Telescope (SPT) survey1 (Ruhl et al., 2004) will cover approximately 4000 deg2 in the mm-bands centered at 150 GHz, 219 GHz, and 274 GHz. The 10 m off-axis Gregorian telescope, shown in the left panel of Fig. 12.1, has seen first light in February 2007 and is currently preparing for routine survey observations. The SPT survey is of special importance for the XMM-Newton Distant Cluster Project because of the substantial overlap of more than 100 XDCP core sample fields with a total nominal exposure time of 3.3 Msec. The ongoing efforts to take full advantage of the joint SZE-X-ray-optical data synergies have been introduced in Sect. 9.2. The future cross-comparison of the X-ray and

1The SPT homepage can be found at http://spt.uchicago.edu/.

223 224 CHAPTER 12. THE FUTURE OF GALAXY CLUSTER SURVEYS

Figure 12.1: Instruments for upcoming major cluster surveys: Left: The 10 m South Pole Telescope has started SZE observations. Image from press release.2Right: eROSITA with its seven mirror and detector systems will conduct the next generation all-sky X-ray survey starting in 2011. Technical drawing from the mission deﬁnition document.3

SZE cluster selection functions will be of particular significance to understand the different survey systematics, which is crucial for the cosmological interpretation of the results. This last important point for the use of galaxy clusters for precision cosmology gave rise to the motivation to conduct a designated X-ray cluster survey (PI: H. Böhringer) in the joint sky regions of the three major SZE instruments SPT, the Atacama Pathfinder EXperiment (APEX-SZ), and the Atacama Cosmology Telescope (ACT). The initial 1 Msec XMM proposal for a 12.5deg2 X-ray coverage with typical exposure times of 12 ksec has been 50% granted at present. XMM observations of the first 42 pointings are currently being conducted. The upcoming X-ray analysis of this cluster survey will largely benefit from the gained experience and developed tools of the XDCP survey.

2The image can be found at http://www.nsf.gov/news/mmg/media/images/SPT%20complete%20s3. JPG. 3The eROSITA mission deﬁnition document can be accessed at http://www.mpe.mpg.de/erosita/ MDD-6.pdf. 12.2. EROSITA 225

12.2 eROSITA

The next generation all-sky X-ray surveys will be conducted with eROSITA (extended ROentgen Survey with an Imaging Telescope Array) starting in 2011. The eROSITA instrument (Predehl et al., 2006) is fully funded and is currently in an active preparation phase as one of the three main experiments of the Russian Spectrum-RG mission. As shown in the right panel of Fig. 12.1, eROSITA will combine seven Wolter-I X-ray mirror modules, each equipped with an improved PN-like detector array with a field-of- view4 of about 0.5deg2. The satellite will be placed in an equatorial Low Earth Orbit with expected good X-ray background properties. The total effective area of 2 500 cm2 is comparable to the imaging mode capabilities of XMM-Newton, but the wider∼ FoV leads to a total grasp of g 700 cm2 deg2, i.e. a survey efficiency improvement of a factor of 2.5 ≃ with respect to the full XMM field or a factor of 3.5 when restricting the XMM FoV to the inner 12 arcminutes. The design specifications concerning the PSF quality aim at a FWHM resolution of 15′′ on-axis and a 20′′ field-of-view average. The current eROSITA survey plans foresee an all-sky survey with a typical extended source sensitivity of 3 10−14 erg s−1cm−2 and a deeper survey region covering a few hundred square degrees to× a flux limit of about 1 10−14 erg s−1cm−2. The challenging× main goal of eROSITA will be the identification of 50000–100000 galaxy clusters out to z 1.5. This unprecedented sample size is required to achieve the ∼ prime science objective of tracing the cosmic evolution of the Dark Energy equation-of-state parameter w using the baryonic acoustic oscillation method discussed in Chap. 3. XDCP will serve as an important pathfinder sample for eROSITA and its distant cluster cosmology applications. As probably the deepest XMM-Newton cluster survey with a sky coverage of 10 deg2 and an average sensitivity 3–5 times better than the eROSITA all-sky ≫ survey, the XDCP sample will be valuable for quantifying the eROSITA performance and for calibrating the cluster selection function. In particular, the lower eROSITA resolution will constrain the accessible parameter region for a distant cluster selection based on extent and is hence to be closely investigated (see Fig. 9.3). In any case, eROSITA will increase the known X-ray cluster population by at least one order of magnitude and will open unprecedented possibilities for precision cosmology with galaxy clusters.

4The eROSITA detector design has recently been modiﬁed and now foresees a FoV diameter of about 60 arcminutes, which will also increase the stated total grasp. 226 CHAPTER 12. THE FUTURE OF GALAXY CLUSTER SURVEYS

ID Method Test Objects σw zmax X O/IR Hi Res 1 BAOs in galaxy cluster PS 1.75 105 GC <5% 2 × × × 2 cosmic shear 109 G <5% 2 3 BAOs in galaxy PS 109 G <5% 3 × × × 4 cluster number counts 1.75 105 GC <10% 2 5 BAOs in AGN PS 5 10×6 AGN <10% 4 × × × × × all methods combined <1% ∼ × × × Table 12.1: Expected results and measurement precision on w of a five year VADER survey covering 3 500 deg2 and probing a comoving volume of 50 Gpc3 for all Dark Energy tests (z 2). ≤ From left to right the table lists the scientific priority (ID), the method, the object class with the expected sample size, and the covered redshift range. The last three columns specify the instrumental requirements for the given methods distinguishing X-ray coverage (X), multi-band optical/IR data (O/IR), and diffraction limited high spatial resolution (Hi Res). By combining all independent VADER measurements of w with their different degeneracies, a sub-percent level precision is achievable.

12.3 VADER – A Concept Study

The ﬁrst two sections covered new galaxy cluster experiments already existing or in preparation. In this ﬁnal section, a concept study is presented, which is closing in on an ultimate cluster survey, i.e. a mission that is able to identify essentially all observable galaxy clusters over a wide sky area. The result of this study is the Very Ambitious Dark Energy Research mission (VADER), a project that originated at the 2005 ESA summer school in Alpbach, Austria. The objective was a design study for a designated Dark Energy satellite mission within ESA’s Horizons 2015–2025 program. The VADER project was continued after the school and led to a publication in the high energy mission section of the SPIE proceedings (Fassbender et al., 2006).

12.3.1 Science case The nature of Dark Energy, as the driving force of the observed acceleration of the Universe, is currently one of the deepest mysteries in astrophysics. A precise knowledge of the equation-of-state parameter w (see Equ. 3.6) and in particular its redshift dependance is required to allow at least a phenomenological understanding of the dominant energy contribution in the Universe. The answer to the fundamental question whether DE is equivalent to Einstein’s cosmological constant (w = 1), phantom energy (Caldwell et al., − 2003) (w< 1), or time varying as in quintessence models (Wetterich, 1988) ( 1

tended X-ray sources, but the science applications depend critically on the time-consuming follow-up observation programs (see Chaps. 7 & 8). VADER is designed as a multi-wavelength survey mission joining X-ray, optical, and IR instruments for a simultaneous spectral coverage from 4 µm (0.3 eV) to 10 keV over a field-of-view of 1 square degree. This combination enables several independent Dark En- ergy tests with low systematics. Table 12.1 lists the five DE tests that were identified with highest priority. The associated expected object sample sizes for galaxy clusters (GC), galaxies (G), and Active Galactic Nuclei (AGN) are specified together with conservative estimates of the measurement accuracy, the covered redshift range, and the instrumental requirements. As the presumably cleanest Dark Energy test (see Sect. 3.2.2), a strong focus is placed on the accurate measurement of the baryonic acoustic oscillations in the power spectrum (PS). This cosmological test (see Tab. 3.2) effectively determines the Hubble expansion history H(z) and angular size distance dang(z) and can be performed independently for the different object classes of galaxy clusters (highest priority), galaxies (medium priority), and AGN (lower priority). Large area measurements of the cosmic shear, i.e. the weak lensing effect of the large-scale structure, is a promising alternative DE technique through its sensitivity to cosmic distances (see e.g. Schneider, 2006b). The determination of the number density evolution of galaxy clusters (Sect. 3.2) completes the main VADER portfolio of Dark Energy tests. Note that the BAO test, cosmic shear, and the cluster number density evolution have later been identified as three of the four most promising routes to Dark Energy research by the NASA Dark Energy Task Force5. Supernovae Ia studies, as the fourth recommended pillar, apply the Hubble diagram method (see Sect. 3.3) but require alternative observational techniques.

12.3.2 Satellite design concept In short, the VADER design foresees the combination of two high-throughput XMM- Newton type X-ray imaging telescopes (0.1–10 keV) complemented by a 1.5 m wide-field corrected optical/IR telescope as sketched in the left panel of Fig.12.2. The optical/IR beam is dichroic-separated6 into eight bands (U G R Z J H K L) allowing the simultaneous coverage from 0.3–4 µm. The two X-ray and eight optical/IR imaging cameras all have a field-of-view of 1 square degree and survey the same patch of sky. As indicated in Tab. 12.1, the weak lensing technique requires high spatial resolution and good optical quality for galaxy shape measurements. Diffraction-limited images are achieved in the Z-band using a 3.7 Giga pixel optical camera with a scale of 0.064′′ per pixel. Most of the other seven camera systems will yield undersampled images with roughly a doubled pixel scale and provide the photometric data for accurate photometric redshifts, the critical pre- requisite of all DE methods. With the proposed eight photometric bands, the achievable expected redshift uncertainties are about ∆z 0.02(1+z) per galaxy and ∆z < 0.01(1+z) ≃ for the averaged cluster redshifts. The near-infrared bands J, H, K, and the∼ additional

5http://www.science.doe.gov/hep/DETF-FinalRptJune30,2006.pdf 6The principal design is similar to the GROND camera. More detailed information can be found at http://www.mpe.mpg.de/∼jcg/GROND. 228 CHAPTER 12. THE FUTURE OF GALAXY CLUSTER SURVEYS

mid-infrared L-band at 3.6 µm provide sensitivities which are inaccessible from the ground and can trace the bulk of the galaxy population out to redshifts of z 3 and for AGN even ∼ further. Additional X-ray imaging data is crucial for the three DE tests using galaxy clusters and AGN as cosmological probes (Tab. 12.1). The two VADER X-ray mirror modules are based on the XMM-Newton design in order to combine the total effective area of 2 800 cm2 at 1.5 keV with an average spatial resolution of about 15′′ across the 1 square degree FoV. This largely extended field-of-view, compared to XMM as discussed in Sect. 6.1, can be achieved with a detector configuration that follows the curved focal plane of the grazing incident mirrors leading to a significant improvement of the off-axis PSF. Including vignetting effects, VADER’s total effective X-ray grasp at 1.5 keV is about 1120 cm2deg2, i.e. a factor of four better than XMM.

12.3.3 Survey and results VADER is designed to perform a very deep extragalactic multi-wavelength survey over a contiguous sky area of 3 500 square degrees during the first five years of the mission. The selected survey region between the South Galactic Pole (SGP) and the South Ecliptic Pole (SEP) is illustrated in the right panel of Fig. 12.2, the possible survey extension and the location of an ultra-deep field are also indicated. The satellite will operate in a highly elliptical 72 hour orbit allowing 60 hours of con- secutive science data acquisition outside the radiation belts. The orbit was selected to optimize the observational and technical requirements of (i) a low X-ray, optical, and IR background, (ii) a high science data fraction per orbit, (iii) long, pointed observations, (iv) observations of a single, contiguous survey field, (v) a stable thermal environment, and (vi) a high downlink rate for data transmission. The survey strategy is aiming for a gross exposure time of 10 h per field, achieved through an observation pattern of overlapping single 2 h exposures sampled over a period of a few weeks in order to additionally allow time domain astrophysics. Taking overheads and data losses into account, the average achievable X-ray flux limit will be about 1 10−15 erg s−1cm−2 for point sources and 2 10−15 erg s−1cm−2 for extended objects. With× × these sensitivities, the expected X-ray selected samples will include approximately 5 106 AGN out to redshifts of at least z 4 and 175000 galaxy clusters out to z 2. These X-ray× capabilities, complemented by the∼ deep optical and infrared galaxy photometry,∼ provide accessability to essentially all existing clusters of galaxies within the survey region with masses of M >1.5 1014 M . The typical achievable limiting AB magnitudes of 26–27 mag × ⊙ in the optical∼ bands and 23–25 mag in the infrared bands will allow the detection, SED characterization, and photometric redshift determination of about one billion galaxies. The special strength of the proposed VADER mission is its ability to provide the most complete and versatile survey data set for Dark Energy and general astrophysical studies. The independent Dark Energy tests using BAOs, cosmic shear, and the cluster number density evolution enable critical cross-checks between the methods. The final combination of all DE techniques with their different degeneracies brings w constraints with percent- level precision within reach. 12.3. VADER – A CONCEPT STUDY 229

Figure 12.2: VADER spacecraft and survey region. Left: Scheme of the VADER spacecraft with its optical/IR telescope and the two X-ray mirror modules. Right: The solid square indicates the 3 500 deg2 sky area between the South Galactic Pole (SGP) and the South Ecliptic Pole (SEP) covered by a 5 year survey. The dashed box shows the prospects for an extended 10 year survey and the small square the location of the proposed 100 deg2 deep survey ﬁeld.

Concerning the technical feasibility of the mission, the X-ray part is mainly based on available technology and bears low risk. The wide-field optical/IR telescope with its multiplexed imaging capability will require a challenging optical design with substantial additional research and development activities. With the rapidly advancing detector technology, the first Giga pixel camera systems will soon be available and similar sizes in the near-infrared regime will become feasible within the next few years. The project scope for a realization of VADER would be an ESA cornerstone mission with a possible launch around 2020. The aim of unveiling the very nature of the dominant Dark Energy component in the Universe has given rise to a hyper-active phase of proposals and developments of a full armada of DE experiments with the result that Dark Energy has become the buzzword in current astrophysics. However, the theoretical and observational developments in Dark Energy research over the next two decades are difficult to foresee and critical opinions on the merit of designated DE missions have appeared (e.g. White, 2007). In any case, the solution to the Dark Energy riddle is of fundamental importance and the route to its understanding might bear many surprises and dead ends. 230 CHAPTER 12. THE FUTURE OF GALAXY CLUSTER SURVEYS

12.4 Concluding Remarks

When speaking of the Golden Age for Astronomy and the dawning Era of Precision Cosmology, one can, without exaggeration, also conceive the next 10–15 years as the New Age of Galaxy Cluster Surveys. Here we have only touched on the upcoming developments in X-rays and for the SZ effect. Major survey projects with a strong focus on galaxy clusters are additionally ongoing or in preparation in the optical, near-infrared, and mid-infrared wavebands. Just to name a few, the Dark Energy Survey (DES), the Kilo Degree Survey (KIDS), the Panoramic Survey Telescope and Rapid Response System (PanSTARRS) are optical experiments with expected cluster sample sizes of at least a few thousand systems. At longer wavelength the Visible & Infrared Survey Telescope for Astronomy7 (VISTA) or the Spitzer mid-infrared fields8 aim at the compilation of hundreds of objects out to high redshift. The common aim of all projects is the identification of systems out to redshifts of z >1, owing to the recognition of galaxy clusters as prime tracers of cosmic evolution. As a survey∼ in an advanced state, the XMM-Newton Distant Cluster Project is in a good position to take on a pathfinder role to largely unexplored redshift regimes over the next few years. In conclusion, the future for distant galaxy cluster studies looks bright. Only the tip of the distant cluster iceberg has been found, the main body is yet to be unveiled.

7A list of VISTA surveys can be found at http://www.eso.org/sci/observing/policies/ PublicSurveys/sciencePublicSurveys.html#VISTA. 8A list of Spitzer surveys is available at http://ssc.spitzer.caltech.edu/legacy/all.html. Appendix A

XDCP Survey Field List

Table A.1: XDCP survey field summary. For each of the 546 processed XMM-Newton observations the central equatorial (RA, DEC) and galactic coordinates (GL, GB) are given followed by the corre- 21 −2 sponding galactic hydrogen column density NH in units of 10 cm . The last five columns contain the XMM field identifier (OBSID), the nominal exposure time (EXT), the effective clean exposure time (CLT), the processing status (STAT), and the target name of the observation (TARGET). 469 fields have been successfully processed and analyzed (STAT=ok), 48 were highly flared (STAT=flared), and 29 were discarded for other reasons (STAT=discard).

RA DEC GL GB NH OBSID EXT CLT STAT TARGET hms dms d.d d.d 1021 sec sec × 00 00 29.2 -25 07 20 40.0 -78.3 0.157 0125310101 46065 16360 ok Abell 2690 00 02 51.2 -29 59 11 15.0 -78.9 0.131 0041750101 52270 46458 ok BLANCO1 00 03 09.1 -35 57 00 349.2 -76.4 0.113 0145020201 54258 47354 ok A 2717 00 03 26.1 -26 02 28 36.0 -79.2 0.163 0103060301 50537 32902 ok Q0000-263 00 00 23.6 +02 04 39 98.0 -58.3 0.300 0201900101 29907 22485 ok RXCJ0003.8+0203 00 10 30.5 +10 58 43 106.9 -50.6 0.576 0127110201 16311 2625 ok IIIZW2 00 11 03.5 -12 07 18 88.7 -72.2 0.257 0150480501 22199 10960 ok NGC34 00 11 25.1 -11 28 42 90.0 -71.7 0.268 0111970901 12857 9673 ok WW Cet 00 14 19.0 -30 22 22 9.0 -81.2 0.165 0042340101 18463 12311 ok RXCJ0014.3-3022 00 14 34.0 -39 10 56 333.1 -75.7 0.176 0028740101 31669 25236 ok NGC 55 (West) 00 15 47.2 -39 15 34 332.1 -75.7 0.176 0028740201 33767 29338 ok NGC 55 (East) 00 18 33.6 +16 26 07 111.6 -45.7 0.407 0111000101 37949 26769 ok CL 0016+16 00 20 45.6 -25 41 43 42.9 -82.9 0.226 0201900301 31007 12852 ok RXCJ0020.7-2542 00 21 18.5 -48 39 13 316.1 -67.6 0.262 0152330101 47961 21255 ok SCG 0018-4854 00 26 07.5 +10 41 11 112.8 -51.6 0.498 0001930101 26609 14168 ok IRAS F00235+10 00 26 36.6 +17 09 36 114.4 -45.3 0.431 0050140201 55009 41795 ok Cl0024+17 00 30 26.7 +04 51 49 113.1 -57.6 0.292 0112320101 31229 8492 ok PSR J0030+0451 00 33 53.0 -43 18 00 314.2 -73.4 0.293 0148961201 19863 15344 ok ELAIS-S1-C

231 232 APPENDIX A. XDCP SURVEY FIELD LIST

00 33 52.9 -43 18 00 314.2 -73.4 0.293 0148961301 48419 32683 ok ELAIS-S1-C 00 33 53.6 -12 07 48 106.7 -74.4 0.242 0125920201 50809 7920 ok G158-100 00 33 52.8 -43 18 00 314.2 -73.4 0.293 0148961401 25039 3379 flared ELAIS-S1-C 00 34 16.1 -21 25 19 87.5 -83.0 0.156 0044350101 22870 11870 ok IRAS00317-2142 00 34 24.6 -05 33 40 111.5 -68.0 0.432 0031740101 35738 7057 ok PSR J0034-0534 00 35 51.3 -43 17 55 312.9 -73.5 0.263 0148960101 104416 47233 ok ELAIS-S1-A 00 37 07.0 +09 09 07 116.9 -53.5 0.478 0084230201 30207 22255 ok Abell 68 00 41 42.2 -09 22 26 115.0 -72.0 0.308 0065140101 13253 9026 discard Abell 85 00 42 30.8 -09 41 25 115.6 -72.4 0.308 0065140201 13248 10324 ok Abell 85 00 43 20.1 +00 51 05 118.6 -61.9 0.245 0090070201 21249 17844 ok UM 269 00 43 24.8 -20 37 31 106.7 -83.2 0.154 0042340201 15006 8511 ok RXCJ0043.4-2037 00 43 35.7 -17 59 25 111.3 -80.6 0.170 0112880601 13250 9422 ok HR 188 00 46 51.5 -20 43 15 113.4 -83.5 0.157 0110990301 16889 2953 flared NGC247 00 47 23.9 -25 15 25 96.8 -87.9 0.149 0110900101 33732 21384 ok NGC 253 NW 00 47 36.0 -25 17 52 97.5 -87.9 0.128 0125960201 17537 5496 ok NGC253 00 47 36.2 -25 17 52 97.5 -87.9 0.128 0125960101 60809 31583 ok NGC253 00 47 39.3 -25 17 08 98.0 -87.9 0.128 0152020101 140799 63718 ok NGC 253 00 49 59.2 -52 08 58 303.4 -64.9 0.340 0133120301 12022 8107 ok BPM 16274 00 50 01.3 -52 08 11 303.4 -64.9 0.340 0123920101 57464 8833 ok BPM 16274 00 50 03.0 -52 08 19 303.4 -64.9 0.340 0125320401 33728 17775 ok BPM16274 00 50 03.3 -52 08 17 303.4 -64.9 0.340 0125320701 45951 10346 ok BPM16274 00 50 07.1 -52 09 41 303.4 -64.9 0.340 0133120401 13707 8611 ok BPM 16274 00 54 55.5 -37 41 09 299.1 -79.4 0.361 0112800101 46711 38155 ok NGC 300 00 54 55.5 -37 41 08 299.1 -79.4 0.361 0112800201 36909 27667 ok NGC 300 00 56 15.0 -01 18 39 125.6 -64.1 0.310 0012440101 34992 25332 discard LBQS0053-0134 00 58 37.0 -36 06 02 293.7 -80.8 0.194 0102040701 21536 7855 ok Q 0056-363 00 58 48.8 -28 25 00 251.6 -87.9 0.195 0111282301 14075 650 flared SGP-9 00 58 53.8 -27 59 00 240.7 -88.1 0.195 0111280601 10944 7947 ok SGP-6 01 02 41.4 -21 52 51 149.5 -84.1 0.160 0144310101 33917 17992 ok Abell 133 01 03 59.4 -06 41 53 131.8 -69.3 0.506 0112650501 25035 16432 ok G133-69 Pos 2 01 04 24.6 -06 24 10 131.9 -69.0 0.508 0112650401 27145 18405 ok G133-69 Pos 1 01 06 22.2 +00 49 11 130.8 -61.8 0.316 0150870201 28815 715 flared BRI0103+0032 01 06 55.9 -80 17 22 302.1 -36.8 0.664 0145540101 31315 22348 ok IRASF01063-8034 01 07 45.8 -17 30 14 145.1 -79.6 0.140 0025540101 12572 3361 flared IC1623 01 12 53.4 -45 32 04 291.2 -71.1 0.210 0067170101 47870 39004 ok Phoenix field 01 13 49.9 -14 50 40 147.0 -76.6 0.163 0147920101 26915 20479 ok MRK 1152 01 18 56.5 -00 59 16 138.2 -63.0 0.371 0153170101 22204 8626 ok MS 0116.3-0115 01 20 20.8 -44 07 51 285.9 -72.0 0.241 0103860901 22608 18679 ok ESO 244- G 017 01 20 39.0 -10 56 27 147.3 -72.4 0.353 0113040801 10768 3539 flared C2001 01 22 43.0 -11 14 24 149.2 -72.4 0.353 0113040701 10045 2455 flared C2001 01 23 46.1 -58 48 19 295.0 -57.8 0.282 0101040201 33006 26928 ok Fairall 9 01 24 35.7 +03 47 20 138.7 -58.0 0.337 0025541601 13203 10287 ok NGC520 01 25 33.4 +01 45 35 140.1 -59.9 0.308 0109860101 41694 33368 ok A 189 233

01 25 47.3 -01 23 55 142.1 -62.9 0.356 0136340101 22515 16496 ok Abell 194 01 31 54.3 -13 36 58 159.9 -73.5 0.156 0084230301 24709 14485 ok Abell 209 01 33 01.1 -40 06 27 272.0 -74.4 0.201 0112630201 37870 21297 ok QSO 0130-403 01 36 20.8 +15 44 24 138.5 -45.7 0.483 0154350201 24914 22087 ok SN2002ap 01 36 24.3 +15 45 02 138.5 -45.7 0.483 0154350101 37259 26902 ok SN2002ap 01 39 52.9 +06 18 31 144.0 -54.5 0.371 0110890901 27412 14774 ok PHL 1092 01 40 10.5 -67 48 45 296.0 -48.6 0.235 0032140401 12561 5279 ok 1saxj0140.2-6748 01 40 55.2 -67 54 28 295.9 -48.5 0.235 0148440101 29559 26028 ok BL Hyi 01 43 01.9 +13 38 22 141.6 -47.3 0.476 0093641001 12209 8034 ok NGC 660 01 46 14.8 -39 22 05 263.4 -73.2 0.186 0090070101 33020 29584 ok Q 0144-3938 01 48 59.9 +05 55 10 147.8 -54.1 0.445 0112551501 21446 14748 ok ngc676 01 52 42.1 +01 00 30 153.0 -58.2 0.284 0084230401 30199 13453 ok Abell 267 01 52 40.0 -13 58 52 173.3 -70.5 0.147 0109540101 53917 42618 ok WARP J0152.7-13 01 53 03.9 -13 43 25 173.0 -70.3 0.169 0112300101 47478 18978 ok ngc 720 01 53 37.2 -49 36 53 279.1 -64.5 0.241 0092970201 14368 10001 ok ESO 197-G010 01 55 46.2 +00 28 58 154.7 -58.3 0.284 0145450201 12618 2216 flared SDSS015543+00 01 59 49.6 +00 23 52 156.5 -57.9 0.251 0101640201 14778 4169 flared Mrk 1014 02 07 39.4 +02 08 10 158.0 -55.4 0.311 0052140301 35006 26292 ok HCG 15 02 07 52.8 +02 43 46 157.6 -54.9 0.350 0048740101 47340 9723 ok NAB 0205+024 02 09 51.5 -63 18 38 288.8 -51.6 0.294 0111970401 15197 6033 ok WX Hyi 02 15 13.1 -73 59 08 295.3 -41.8 0.339 0049150101 29614 5694 ok Magellanic Bridge 02 15 19.6 -73 59 50 295.3 -41.8 0.339 0099840101 14256 9602 ok Magellanic Bridge 02 15 27.2 -74 01 09 295.3 -41.8 0.339 0049150201 28609 4147 flared Magellanic Bridge 02 15 35.6 -73 59 57 295.2 -41.8 0.339 0049150301 31426 15332 ok Magellanic Bridge 02 19 23.7 -02 57 50 167.7 -57.9 0.244 0148500201 12218 10172 ok Mira 02 22 00.2 -04 50 12 171.0 -58.9 0.255 0112680801 15617 10372 ok MLS-4 02 24 03.2 -04 29 12 171.3 -58.3 0.251 0112680501 23645 17817 ok MLS-8 02 24 36.9 -04 10 51 171.1 -58.0 0.232 0112680301 23411 19205 ok MLS-3 02 25 08.0 +18 47 37 151.7 -38.7 1.042 0150180101 22208 19599 ok RBS 315 02 25 13.3 -29 27 30 224.9 -69.2 0.170 0201900701 29481 0 flared RXCJ0225.1-2928 02 25 20.5 -04 30 10 171.7 -58.1 0.232 0112681001 41946 19995 ok MLS-7 02 26 02.8 -04 09 10 171.5 -57.7 0.232 0112680201 21648 8058 ok MLS-2 02 26 43.0 -04 29 13 172.2 -57.9 0.232 0112681301 40381 13322 ok MLS-6 02 27 20.6 -04 10 11 172.0 -57.5 0.273 0112680101 30351 22689 ok MLS-1 02 28 00.4 -04 30 10 172.7 -57.7 0.273 0112680401 25050 20187 ok MLS-5 02 28 56.1 -10 10 25 181.1 -61.4 0.260 0112860101 10035 4697 ok Q0226-1024 02 32 21.0 -44 19 45 259.9 -63.4 0.261 0042340301 13645 9640 ok RXCJ0232.2-4420 02 34 22.6 -43 47 43 258.4 -63.4 0.261 0148790101 48479 19176 ok CC Eri 02 34 41.4 -08 46 35 180.8 -59.4 0.283 0150470601 57917 12453 ok NGC 985 02 36 11.5 -52 19 16 272.2 -58.1 0.308 0098810101 24995 20509 ok WW Hor 02 38 19.4 -52 11 38 271.6 -57.9 0.307 0067190101 34146 18519 ok ESO 198-G24 02 38 38.5 +16 36 49 156.7 -39.1 0.885 0110990101 19851 16422 ok AO0235+164 02 41 04.4 -08 15 12 182.0 -57.9 0.308 0093630101 16463 12911 ok NGC 1052 234 APPENDIX A. XDCP SURVEY FIELD LIST

02 42 40.9 -00 00 47 172.1 -51.9 0.355 0111200101 42258 33615 ok NGC 1068 02 42 41.0 -00 00 46 172.1 -51.9 0.355 0111200201 46429 28657 ok NGC 1068 02 48 06.6 -03 30 57 177.7 -53.4 0.412 0084230501 33645 25182 ok Abell 383 02 49 33.7 -31 11 24 229.0 -63.9 0.227 0146510401 38914 29923 ok IC1860 Group 02 52 54.5 -01 15 39 176.4 -51.0 0.525 0151490101 47204 19428 ok NGC 1132 02 56 05.6 +19 22 48 159.2 -34.5 1.065 0112510301 30242 20763 ok MBM 12 Pos 2 02 56 04.5 +19 28 50 159.2 -34.4 1.096 0110660101 24196 10395 ok MBM 12 02 56 33.4 +00 06 00 175.9 -49.4 0.665 0056020301 25094 8676 ok RX J0256.5+0006 02 57 52.5 +13 02 08 164.3 -39.4 1.058 0112260101 15409 9110 ok A399 02 58 58.4 +13 33 47 164.2 -38.8 1.019 0112260301 13806 9988 ok A399 02 59 51.4 +13 46 47 164.2 -38.5 1.019 0112260401 12755 8971 ok A399 03 02 39.1 +00 07 31 177.5 -48.3 0.733 0041170101 51724 40890 ok CFRS3h 03 05 22.2 +17 29 11 162.7 -34.8 1.106 0112190101 13647 9886 ok MS0302.5+1717 03 06 37.6 +00 00 28 178.6 -47.6 0.668 0142610101 73918 32406 ok S2F1a 03 07 04.4 -28 40 29 223.9 -60.0 0.136 0042340501 14767 10108 ok RXCJ0307.0-2840 03 11 55.8 -76 51 52 293.4 -37.5 0.748 0122520201 46099 23563 ok PKS0312-770 03 17 56.7 -44 14 10 252.9 -56.0 0.253 0105660101 24470 19260 ok A3112 03 18 14.9 -66 29 48 283.3 -44.6 0.440 0106860101 42416 15312 ok NGC 1313 03 19 59.4 +11 14 10 171.1 -37.3 1.671 0110661101 12796 8587 ok MBM 16 03 20 07.6 -62 26 33 278.6 -47.0 0.422 0084960101 27842 6428 ok Galactic Halo 03 23 21.1 -49 30 26 260.9 -53.3 0.172 0140190101 29614 25722 ok RX J0323-4931 03 23 50.5 -37 17 11 240.2 -56.4 0.167 0091770101 60362 27252 ok NGC 1316 03 32 53.2 -09 28 23 195.8 -48.0 0.413 0112880501 13420 11301 ok HR 1084 03 32 52.8 -63 27 53 278.6 -45.3 0.504 0084960201 12846 10014 ok Galactic Halo 03 33 30.3 -36 10 22 238.0 -54.6 0.135 0151370101 19417 15229 ok NGC 1365-X1 03 33 38.8 -36 08 55 237.9 -54.5 0.135 0151370701 11417 6839 ok NGC 1365-X1 03 36 41.1 +00 36 23 184.8 -41.5 0.805 0116150601 64197 22191 ok HR1099 03 36 41.3 +00 36 24 184.8 -41.5 0.805 0116340601 28610 17959 discard HR1099 03 36 43.7 -36 01 03 237.6 -53.9 0.139 0140950201 17417 14528 ok NGC1386 03 36 47.2 +00 35 21 184.9 -41.5 0.805 0134540601 35740 28264 ok HR1099 03 36 46.9 +00 35 27 184.9 -41.5 0.805 0129350201 31310 8460 ﬂared HR1099 03 36 47.0 +00 35 20 184.9 -41.5 0.805 0134540101 46325 13660 discard HR1099 03 36 47.2 +00 35 19 184.9 -41.5 0.805 0134540801 51927 27 discard HR1099 03 37 44.3 -25 22 17 219.8 -52.8 0.155 0107860401 63138 26483 ok SHARC4 03 38 28.3 -35 26 52 236.7 -53.6 0.145 0012830101 29364 6170 ok NGC 1399 03 38 29.8 +00 21 44 185.5 -41.3 0.735 0036540101 23004 17982 ok SDSS033829+00 03 38 36.3 +09 57 50 176.2 -35.0 1.864 0109870101 30612 9118 discard 2A 0335+096 03 38 40.7 +09 58 17 176.2 -35.0 1.864 0147800201 140323 99140 discard 2A 0335+096 03 39 35.7 -35 25 55 236.6 -53.4 0.148 0055140101 51410 39957 ok LP 944-20 03 41 30.2 -44 52 34 252.0 -51.8 0.166 0045940301 31893 22907 ok AXJ0341.4-4453 03 45 46.0 -41 12 39 246.0 -51.7 0.188 0201900801 26912 12926 ok RXCJ0345.7-4112 03 53 43.6 -00 04 33 188.9 -38.5 1.183 0134920901 20467 9150 ok SA95-42 03 53 50.5 -10 24 35 200.6 -44.0 0.428 0093160201 28014 13589 ok BR 0351-1034 235

03 54 35.6 -20 30 52 214.1 -47.8 0.394 0150910101 18920 7796 ok NGC 1482 03 57 25.3 +01 11 41 188.3 -37.0 1.301 0094790201 21892 19221 ok Hawaii 167 03 58 50.8 +10 25 21 179.8 -31.0 1.452 0064600301 11914 6857 ok 3C 98 03 58 57.9 +10 26 37 179.8 -31.0 1.452 0064600101 28904 9336 ok 3C 98 04 12 35.5 +10 15 34 182.4 -28.5 1.432 0112231601 38757 17000 ok A478 oﬀ-set 04 13 30.2 +10 28 05 182.4 -28.2 1.527 0109880101 126748 53202 ok A 478 04 19 34.3 +02 23 54 190.9 -31.8 1.159 0152150101 30912 19525 ok NGC1550 04 19 33.2 +14 32 50 179.9 -24.5 1.470 0141400301 25218 8165 ok LH0416+14 04 19 47.6 +15 37 33 179.0 -23.8 1.764 0024140101 63578 45054 ok GammaTau 04 20 01.1 -50 23 54 258.1 -44.3 0.196 0141750101 22201 19269 ok RXJ0420.0-5022 04 20 01.0 -50 23 55 258.1 -44.3 0.196 0141751001 21914 16225 ok RXJ0420.0-5022 04 20 00.7 -50 23 56 258.1 -44.3 0.196 0141751101 22420 19204 ok RXJ0420.0-5022 04 20 05.5 -50 21 47 258.1 -44.3 0.196 0141751201 21919 19404 ok RXJ0420.0-5022 04 22 12.6 -38 45 55 241.7 -44.9 0.190 0104860101 22865 8417 ok [HB89] 0420-388 04 24 12.7 +14 45 35 180.5 -23.5 1.670 0101440601 48757 33085 ok VB50 04 24 12.9 +14 45 28 180.5 -23.5 1.670 0101441501 57513 32175 ok VB50 04 26 11.2 +16 55 48 179.0 -21.8 1.915 0056020401 23659 8990 ok RX J0426.1+1655 04 28 35.0 +15 57 45 180.2 -21.9 1.711 0101440501 47811 37399 ok VB71 04 31 38.8 +18 10 15 178.9 -20.0 1.792 0109060301 57818 48640 ok L1551 cloud 04 33 39.3 -13 15 45 209.5 -36.4 0.568 0135120201 30867 17308 discard A 496 04 33 56.5 -08 35 24 204.4 -34.3 0.579 0150480201 23929 17124 ok NGC1614 04 37 14.1 +00 44 13 195.3 -29.0 0.819 0042341601 18238 2955 ﬂared RXCJ0437.1+0043 04 37 22.0 -47 15 30 253.4 -41.9 0.169 0112320201 69412 24942 ok PSR J0437-4715 04 40 16.9 -43 33 25 248.4 -41.5 0.147 0104860201 12868 10136 ok [HB89] 0438-436 04 47 52.1 -06 27 17 204.0 -30.3 0.547 0014740601 28887 6505 ok a0447-0627 04 48 12.6 -20 29 08 219.5 -35.9 0.321 0142240101 15208 7895 ok Abell 514 04 48 30.8 -20 35 42 219.6 -35.8 0.321 0142240201 15203 7860 ok Abell 514 04 52 34.4 -29 53 01 231.1 -37.5 0.180 0153100101 15925 9924 ok IRAS 04505-2958 04 54 10.2 -53 22 40 261.1 -38.7 0.350 0148650101 58916 39557 ok NGC 1705 04 54 44.8 -18 08 30 217.4 -33.6 0.425 0140210101 38947 29209 ok RXCJ0454-1808 04 59 34.8 +01 47 17 197.6 -23.7 0.747 0112880401 20600 16321 ok Gl 182 05 00 55.6 -38 40 12 242.4 -37.3 0.307 0083151201 13516 8438 ok A3301 05 01 05.9 -24 25 18 225.2 -34.3 0.237 0110870101 25867 21755 ok CL0500-24 05 05 19.4 -28 48 53 230.6 -34.6 0.154 0111160201 49616 26291 ok SHARC-2 05 07 42.0 -37 31 00 241.2 -35.9 0.278 0110980801 43606 33560 ok NGC 1808 05 12 55.9 -16 12 25 217.2 -28.9 0.600 0143370101 47265 37618 ok Mu Lep 05 14 06.2 -40 02 43 244.5 -35.0 0.361 0151750101 23914 8177 ok 4U 0513-40 05 15 41.5 +01 04 28 200.4 -20.6 1.091 0010620101 31641 20228 ok V1309 Ori 05 16 43.8 -54 30 26 262.2 -35.3 0.625 0042340701 15114 6620 ok RXCJ0516.7-5430 05 19 49.8 -45 46 59 251.6 -34.6 0.449 0090050701 20521 8647 ok PICTOR A 05 21 01.6 -25 21 58 227.8 -30.2 0.193 0085640101 12423 7693 ok IRAS05189-2524 05 23 03.3 -36 27 22 240.6 -32.6 0.329 0065760201 31919 28588 ok PKS0521-365 05 25 01.7 -33 43 47 237.5 -31.7 0.218 0149500801 13706 6745 ok PMN-525-3343 236 APPENDIX A. XDCP SURVEY FIELD LIST

05 25 01.7 -33 43 46 237.5 -31.7 0.218 0149500901 12205 5727 ok PMN-525-3343 05 25 01.5 -33 43 46 237.5 -31.7 0.218 0149500601 12206 8994 ok PMN-525-3343 05 25 01.7 -33 43 45 237.5 -31.7 0.218 0149500701 12206 8534 ok PMN-525-3343 05 25 06.2 -33 42 51 237.5 -31.6 0.218 0050150301 28528 17343 ok PMN 0525-3343 05 25 06.7 -33 43 12 237.5 -31.6 0.218 0050150101 21870 8102 ok PMN 0525-3343 05 25 08.9 -33 42 09 237.5 -31.6 0.218 0149501201 12419 5663 ok PMN-525-3343 05 28 56.0 -39 27 36 244.3 -32.1 0.210 0042340801 14112 5543 ok RXCJ0528.9-3927 05 33 01.9 -37 01 21 241.7 -30.8 0.293 0042341801 12740 8292 ok RXCJ0532.9-3701 05 33 44.0 -24 10 49 227.6 -27.1 0.233 0068940101 129506 21632 ok CL1 05 33 43.7 -24 10 49 227.6 -27.1 0.233 0068940601 35178 11860 ok CL1 05 40 12.0 -40 54 59 246.5 -30.2 0.353 0149420101 18217 14790 ok ESO03060170 05 47 38.6 -31 52 42 236.9 -26.6 0.195 0201900901 25320 20193 ok RXCJ0547.6-3152 05 50 40.8 -32 16 09 237.5 -26.1 0.220 0111830201 54576 15715 ok PKS 0548-32 05 59 46.1 -50 26 59 257.9 -28.5 0.484 0125110101 57265 8000 ok PKS0558-504 05 59 47.2 -50 26 52 257.9 -28.5 0.484 0129360201 26409 9031 ok PKS0558-504 05 59 46.6 -50 27 07 257.9 -28.5 0.484 0119100201 45437 4340 flared PKS0558 05 59 48.9 -50 26 37 257.9 -28.5 0.484 0137550201 14968 11568 ok PKS0558-504 05 59 48.4 -50 27 10 257.9 -28.5 0.484 0137550601 14785 4640 ok PKS0558-504 06 02 02.8 -39 57 37 246.4 -26.0 0.501 0151900101 47206 19753 ok A3376 06 15 24.3 -33 24 56 240.5 -21.5 0.325 0092360301 15513 6363 ok 3EG 0616-3310 06 15 24.1 -32 55 00 240.0 -21.3 0.410 0092360401 15514 8972 ok 3EG 0616-3310 06 16 47.1 -47 48 12 255.6 -25.3 0.481 0201901101 44676 2156 flared RXCJ0616.8-4748 06 17 47.1 -32 54 57 240.1 -20.9 0.410 0092360101 12512 8503 ok 3EG 0616-3310 06 17 47.1 -33 24 55 240.6 -21.0 0.384 0092360201 15068 10551 ok 3EG 0616-3310 06 45 21.9 -54 12 57 263.6 -22.5 0.657 0201901201 26397 10965 ok RXCJ0645.4-5413 06 45 25.6 -54 12 11 263.6 -22.5 0.657 0201903401 23236 11474 ok RXCJ0645.4-5413 06 58 16.8 -55 57 23 266.0 -21.2 0.634 0112980201 46756 21497 ok RXJ0658-55 07 51 08.8 +18 07 50 202.7 21.0 0.502 0111100301 38499 9271 ok PSR J0751+18 08 06 27.5 +15 27 49 206.9 23.4 0.248 0150800101 28719 17091 ok RXJ0806.3+1527 08 39 57.5 +19 32 39 206.0 32.3 0.309 0101440401 48305 39011 ok Praesepe 08 40 52.2 +13 12 39 212.9 30.1 0.422 0147670301 30914 8263 ok 3C 207 08 51 26.7 +11 47 08 215.7 31.9 0.377 0109461001 10160 7267 ok EU Cnc 09 08 13.8 -09 36 35 239.1 24.6 0.421 0136740201 18366 5363 ok A754 f2 09 08 45.6 -09 46 03 239.3 24.6 0.421 0112950401 16810 9772 ok A754 f4 09 08 55.4 -09 32 04 239.1 24.8 0.421 0112950301 14771 9836 ok A754 f3 09 09 26.7 -09 40 22 239.3 24.8 0.459 0136740101 18370 11654 ok A754 f1 09 11 27.6 +05 51 05 224.6 33.6 0.391 0083240201 20814 11284 discard RX J0911.4+0551 09 18 00.9 -12 05 11 242.9 25.0 0.486 0109980301 32995 14662 ok Hydra A cluster 09 43 44.0 +16 44 34 216.3 45.5 0.282 0046940401 15560 11234 ok WARPJ0943.7+16 09 43 43.8 +16 44 05 216.3 45.5 0.282 0046940201 33116 3149 flared WARPJ0943.7+16 09 45 41.8 -14 19 40 249.7 28.7 0.517 0147920301 28917 19551 ok NGC 2992 09 52 08.6 -01 48 07 239.4 38.0 0.384 0065790101 10111 5800 ok RXJ 095208.7-014 09 53 04.9 +07 55 24 228.8 43.6 0.301 0103260801 84291 15600 ok PSR B0950+08 237

09 53 14.3 -15 58 43 252.5 28.9 0.527 0140210201 38915 33996 ok RXCJ0953-1558 09 53 36.5 +01 34 33 236.1 40.3 0.359 0070940401 26760 7633 ok NGC 3044 09 53 41.0 +01 35 03 236.1 40.3 0.359 0070940101 11428 3713 ﬂared NGC 3044 09 54 48.8 -20 23 03 256.2 26.0 0.389 0141170101 61227 30122 ok EIS0954-2023 09 54 56.7 +17 43 19 216.4 48.3 0.339 0112850101 33376 14111 ok 0952+179 09 56 11.7 -10 03 09 248.0 33.5 0.507 0148170101 102815 64021 ok Abell 901 09 58 17.6 -11 03 29 249.3 33.2 0.506 0201903501 18285 6509 ok RXCJ0958.3-1103 10 00 11.9 -19 38 08 256.7 27.4 0.449 0041180301 22358 18131 ok NGC 3091 group 10 04 15.7 +05 12 53 234.1 44.6 0.233 0150610101 23624 8447 ok PG 1001+054 10 07 21.7 +12 48 35 225.1 49.1 0.369 0140550601 22205 19384 ok PG 1004+130 10 11 05.7 -04 41 20 246.1 39.8 0.391 0026340201 21618 6073 ok Sextans A 10 19 59.8 +08 13 28 233.5 49.5 0.329 0093640301 24263 3845 ﬂared IRAS 10173+0828 10 21 37.4 +13 06 40 227.2 52.3 0.379 0146990101 21917 16357 ok IRAS 10190+1322 10 23 30.8 +19 51 56 216.9 55.4 0.222 0101040301 40495 32296 ok NGC 3227 10 23 40.0 +04 11 27 239.3 47.9 0.287 0108670101 56459 48492 ok ZW 3146 10 28 34.2 -08 44 41 253.9 40.1 0.459 0153290101 43368 15915 ok RXJ1028.6-0844 10 30 22.4 +05 24 45 239.4 50.0 0.339 0148560501 103876 60819 ok SDSS 1030+05 10 36 28.2 -03 43 17 251.1 45.1 0.471 0150870401 31418 27773 ok BRI1033-0327 10 41 12.8 +06 10 10 241.0 52.6 0.270 0151390101 59915 40999 ok 4C 06.41 10 44 32.9 -01 25 13 250.8 48.2 0.432 0125300101 62310 29778 ok J104433.04-0125 10 50 26.2 -12 50 39 262.7 40.4 0.501 0146510301 40921 20572 ok NGC3411 Group 10 56 59.8 -03 37 38 256.5 48.6 0.379 0094800101 41021 22223 ok MS1054.4-0321 11 01 52.6 -34 42 37 278.6 22.9 0.473 0112880201 29776 25669 ok TW Hya 11 06 34.2 -18 21 44 270.8 37.8 0.461 0112630101 36428 10464 ok HE 1104-1805 11 13 17.2 -26 45 06 277.2 31.1 0.561 0071340201 22234 10958 ok NGC 3585 11 15 18.1 -21 56 38 275.2 35.7 0.408 0094380101 33673 26861 ok GRB011211 11 17 15.8 +17 57 42 230.8 66.4 0.143 0099030101 23795 13319 ok DP Leo 11 18 17.3 +07 46 11 249.8 60.6 0.336 0082340101 63358 49529 ok PG1115+080 11 18 55.1 +13 05 19 241.3 64.2 0.227 0082140301 33519 28398 ok NGC3623 11 20 14.9 +12 59 24 241.9 64.4 0.264 0093641101 11363 6979 ok NGC 3627 11 20 17.1 +13 35 31 240.8 64.7 0.196 0110980101 86739 41222 ok NGC 3628 11 23 09.2 +05 30 29 254.7 59.8 0.433 0083000301 33361 24867 ok 3C 257 11 23 16.1 +01 37 42 259.5 56.8 0.423 0145750101 42208 21840 ok Q1120+0195 11 30 02.3 -14 49 25 275.2 43.6 0.406 0112850201 30089 13489 ok 1127-145 11 31 56.5 -19 55 42 278.5 39.1 0.451 0042341001 15025 10122 ok RXCJ1131.9-1955 11 32 00.5 -34 36 30 284.8 25.4 0.504 0112880101 29921 26857 ok CoD-33 7795 11 38 22.5 +03 22 05 263.4 60.5 0.235 0111970701 12866 10213 ok T Leo 11 39 01.7 -37 44 08 287.4 22.9 0.941 0112210201 137818 16000 ok NGC3783 11 39 01.7 -37 44 08 287.4 22.9 0.941 0112210501 137815 20000 ok NGC3783 11 39 02.3 -37 44 13 287.4 22.9 0.941 0112210101 40412 13509 ok NCG3783 11 39 06.9 +17 07 15 240.1 70.5 0.203 0066950201 13383 4620 ok UGC 6614 11 41 07.0 -01 43 09 269.7 56.5 0.245 0151230101 10919 6607 ok UN J1141-0143 11 41 19.7 -12 16 20 277.3 47.0 0.330 0201901601 37932 24496 ok RXCJ1141.4-1216 238 APPENDIX A. XDCP SURVEY FIELD LIST

11 44 39.9 +19 47 00 235.0 73.0 0.255 0061740101 33366 26714 discard A1367 11 50 41.5 +01 45 50 270.5 60.7 0.215 0044740201 49072 43910 ok Beta Vir 11 51 02.4 -28 48 06 287.2 32.2 0.627 0027340101 45533 29905 ok NGC 3923 12 00 48.0 -03 27 39 279.2 57.0 0.243 0056020701 30114 24816 ok RX J1200.8-0328 12 01 53.1 -18 51 42 286.9 42.4 0.400 0085220101 24913 15885 ok The Antennae 12 01 52.6 -18 51 41 286.9 42.4 0.400 0085220201 52777 34943 ok The Antennae 12 04 27.4 +01 54 30 276.9 62.3 0.186 0093060101 16004 11301 ok MKW4 12 12 15.8 +13 12 33 267.6 73.3 0.249 0112550501 23614 17974 ok ngc4168 12 13 46.0 +02 48 53 280.9 64.0 0.174 0081340801 23206 19206 ok IRAS12112+03 12 16 40.5 -12 01 17 289.6 49.9 0.385 0143210801 33916 22276 ok cl1216-12 12 18 45.0 +14 24 53 270.3 75.1 0.271 0147610101 50920 15073 ok NGC 4254 12 19 23.2 +05 49 47 281.8 67.3 0.163 0056340101 33360 23324 ok NGC 4261 12 20 13.5 +06 41 15 281.5 68.2 0.158 0105070101 11482 5419 ok 1WGA J1220.3+06 12 22 54.9 +15 49 19 271.1 76.8 0.232 0106860201 36633 18017 ok NGC 4321 12 23 06.8 +10 37 22 279.5 72.1 0.217 0108860101 22271 17203 ok NGC 4325 12 25 42.4 +12 39 33 279.0 74.3 0.263 0110930301 18760 433 flared NGC 4388 12 25 51.5 +12 39 43 279.1 74.3 0.263 0110930701 11915 8140 ok NGC 4388 12 26 07.3 +12 56 38 279.0 74.6 0.264 0108260201 85636 55860 discard M86 12 27 18.9 +01 29 11 289.2 63.7 0.181 0110990201 29679 9818 ok HI1225+01 12 28 13.7 -15 46 59 294.7 46.7 0.378 0002740301 12234 6450 ok Denis-J1228 12 28 44.2 +02 07 20 289.7 64.4 0.179 0126700401 27251 2696 flared 3C 273off+7min 12 29 03.3 +02 01 42 289.9 64.3 0.179 0126700201 26140 8678 ok 3C 273off-1.5min 12 29 30.2 +01 59 32 290.2 64.3 0.179 0126700101 27248 135 flared 3C 273off-7min 12 29 41.4 +07 59 45 286.8 70.1 0.159 0112550601 24622 11715 discard ngc4472 12 29 57.5 +13 38 11 281.4 75.6 0.280 0112552101 14126 10301 ok ngc4477 12 30 06.3 +13 38 53 281.5 75.6 0.258 0112550701 13488 4252 flared ngc4477 12 30 45.5 +13 43 32 282.0 75.7 0.258 0106060401 13275 8099 ok Virgo 4 12 30 45.4 +14 43 22 280.6 76.6 0.247 0106060701 14375 10637 ok Virgo 7 12 30 45.3 +15 23 29 279.6 77.3 0.247 0106060901 15622 8970 ok Virgo 9 12 30 45.3 +15 43 30 279.0 77.6 0.230 0106061001 15720 6231 ok Virgo 10 12 30 45.2 +16 03 33 278.4 77.9 0.230 0106061101 16624 7389 ok Virgo 11 12 30 45.2 +14 03 18 281.6 76.0 0.258 0106060501 17738 12421 ok Virgo 5 12 30 45.4 +14 23 22 281.1 76.3 0.258 0106060601 14876 8756 ok Virgo 6 12 30 45.2 +16 23 31 277.8 78.2 0.230 0106061201 18441 11858 ok Virgo 12 12 30 45.4 +17 53 28 274.5 79.6 0.251 0106061301 16624 12174 ok Virgo 13 12 30 45.3 +15 03 29 280.1 77.0 0.247 0106060801 13057 1660 flared Virgo 8 12 30 49.1 +12 23 22 283.7 74.4 0.250 0114120101 60109 30143 discard M87 12 30 49.9 +12 43 17 283.4 74.8 0.259 0106060101 10020 4160 flared Virgo 1 12 30 53.9 +11 00 03 285.2 73.1 0.209 0145800101 107002 50080 discard LBQS 1228+1116 12 31 58.3 +14 25 19 282.3 76.5 0.247 0112550801 14163 6669 ok ngc4501 12 35 34.9 -39 54 50 299.6 22.8 0.719 0006220201 46207 35856 ok NGC 4507 12 35 39.7 +12 33 17 287.9 74.9 0.259 0141570101 44838 20074 ok NGC 4552 12 36 39.4 -33 54 03 299.4 28.8 0.561 0201901701 26147 5973 ok RXCJ1236.7-3354 239

12 37 38.8 +11 49 03 290.3 74.3 0.261 0112840101 23669 15453 ok NGC4579 12 39 38.9 -05 20 43 297.4 57.4 0.228 0109970101 28062 0 flared NGC 4593 12 39 59.3 -11 37 11 298.4 51.1 0.385 0084030101 43456 23727 ok M104 12 40 20.4 -83 14 57 302.5 -20.3 0.963 0092360801 16835 11397 ok 3EG 1249-8330 12 40 21.2 -83 44 57 302.6 -20.8 0.860 0092360701 13236 2275 flared 3EG 1249-8330 12 41 13.5 +18 34 32 287.0 81.1 0.170 0149900301 17412 14070 ok Abell 1589 12 42 38.3 -11 19 31 299.4 51.4 0.374 0136950201 30265 26157 ok RXJ1242 12 42 48.2 +02 41 21 297.7 65.4 0.175 0111190701 64406 53805 ok NGC4636 12 42 51.8 +13 15 40 294.2 75.9 0.236 0112551001 15112 10990 ok ngc4639 12 43 39.8 +11 33 15 295.8 74.3 0.194 0021540201 54210 43902 ok NGC 4649 12 45 04.5 -00 27 41 299.5 62.3 0.174 0110980201 58237 50432 ok NGC 4666 12 48 22.0 +08 29 24 300.5 71.3 0.192 0112551101 16912 10529 ok ngc4698 12 48 48.9 -41 18 42 302.4 21.5 0.825 0046340101 47807 30632 discard Centaurus cluster 12 49 13.8 -05 59 34 301.9 56.8 0.212 0060370201 41273 30143 ok Q1246-057 12 52 24.2 -29 15 02 303.1 33.6 0.610 0111020201 36207 6496 ok EX Hydrae 12 52 24.2 -29 15 02 303.1 33.6 0.610 0111020101 57802 0 flared EX Hydrae 12 52 59.3 -29 27 13 303.3 33.4 0.600 0057740301 68621 60049 ok c1252.9-2927 12 52 59.3 -29 27 03 303.3 33.4 0.600 0057740401 68626 60993 ok c1252.9-2927 12 53 10.5 -09 11 56 303.6 53.6 0.293 0112270701 12866 9766 ok HCG62 12 53 37.6 +15 42 24 305.5 78.5 0.191 0082990101 50900 29195 ok 3C 277.2 12 54 00.0 +10 11 25 305.0 73.0 0.162 0001930301 25335 18403 ok IRAS F12514+10 12 54 33.6 -29 08 17 303.7 33.7 0.610 0030140101 17248 13787 discard A3528 12 57 16.8 -30 22 11 304.4 32.4 0.596 0030140301 16895 11650 ok A3532 12 57 28.2 -17 20 29 304.9 45.5 0.465 0010420201 22806 12570 discard Abell 1644 12 58 00.5 -83 14 57 303.1 -20.3 0.710 0092360501 15383 7728 ok 3EG 1249-8330 12 57 59.9 -83 44 59 303.1 -20.8 0.689 0092360601 15380 8747 ok 3EG 1249-8330 12 58 48.0 -01 44 44 306.7 61.0 0.154 0093200101 43240 33950 ok A1650 13 02 46.1 -02 30 29 308.6 60.2 0.171 0201901801 29149 20164 ok RXCJ1302.8-0230 13 05 13.1 -10 20 50 308.4 52.3 0.322 0032141201 16003 12223 ok 1saxj1305.2-1020 13 05 43.9 +18 01 06 323.5 80.3 0.198 0017940101 56505 40187 ok GP Com 13 11 29.3 -01 20 18 313.3 61.1 0.180 0093030101 39806 32201 ok Abell 1689 13 15 23.8 -16 22 55 311.2 46.1 0.491 0037950101 24109 18674 ok NGC 5044 13 15 28.6 -16 45 29 311.1 45.7 0.491 0152360101 39116 34508 ok NGC 5044 13 19 16.5 -14 50 42 312.9 47.4 0.491 0110980601 53647 27626 ok NGC 5073 13 24 05.1 +13 58 28 334.6 74.8 0.175 0108860701 26193 14219 ok NGC 5129 13 25 19.9 -38 24 43 310.1 23.9 0.481 0110890101 64019 52553 ok IRAS 13224-3809 13 27 57.2 -31 30 19 311.9 30.7 0.363 0107260101 44615 37883 discard A3558 13 29 21.7 +11 46 24 334.8 72.2 0.195 0041180801 24948 16594 ok NGC 5171 Group 13 30 53.8 -01 50 15 322.6 59.5 0.232 0112240301 34779 26166 ok A1750 13 31 45.6 -31 48 02 312.8 30.2 0.398 0105261401 23338 10413 ok A3562 f2 13 32 24.9 -31 37 27 313.0 30.4 0.391 0105261601 23847 17065 ok A3562 f4 13 32 55.3 -31 52 05 313.1 30.1 0.398 0105261701 20911 17323 ok A3562 f5 13 33 05.4 -31 41 25 313.1 30.3 0.391 0105261301 47167 36328 ok A3562 f1 240 APPENDIX A. XDCP SURVEY FIELD LIST

13 33 54.7 -31 30 26 313.4 30.4 0.391 0105261501 22648 18919 ok A3562 f3 13 34 25.2 -31 44 20 313.5 30.2 0.391 0105261801 21847 8385 ok A3562 f6 13 35 53.9 -34 17 49 313.2 27.6 0.409 0111570101 46453 8000 ok MCG-6-30-15 13 35 53.9 -34 17 47 313.2 27.6 0.409 0111570201 66197 8000 ok MCG-6-30-15 13 37 05.1 -29 51 46 314.6 31.9 0.397 0110910201 30638 20121 ok M 83 13 37 44.4 -12 57 20 320.0 48.3 0.481 0147670201 13917 11814 ok PKS B1334-127 13 38 11.5 +04 32 24 331.2 64.8 0.197 0152940101 67313 30686 ok NGC 5252 13 39 56.2 -31 38 39 314.8 30.1 0.390 0035940301 47216 29616 ok NGC 5253 13 41 19.7 +00 23 53 329.1 60.7 0.186 0111281001 10377 7333 ok F864-1 13 41 19.6 -00 00 07 328.8 60.3 0.191 0111281301 14541 2342 flared F864-4 13 47 15.5 +17 27 21 358.9 73.8 0.181 0144570101 66431 39983 ok Tau Boo 13 47 22.7 -33 09 54 316.2 28.2 0.430 0147040101 31428 11078 discard Abell 3571 13 47 30.7 -11 45 11 324.0 48.8 0.492 0112960101 38392 30423 ok RXJ1347-1145 13 47 30.1 -32 51 52 316.3 28.5 0.393 0086950201 33642 23354 discard A3571 13 49 19.0 -30 18 30 317.4 30.9 0.443 0101040401 13925 9262 discard IC 4329a 13 50 43.9 -33 43 05 316.8 27.5 0.480 0201901901 26515 15442 ok RXCJ1350.7-3343 13 54 12.6 -02 21 52 332.5 56.8 0.353 0112250101 33646 22728 ok RXJ1354.3-0222 13 56 03.0 +18 22 33 5.8 72.7 0.208 0094401201 26850 22981 ok Mrk 463 14 01 01.8 +02 52 40 340.3 60.5 0.224 0098010101 61353 27531 ok A 1835 14 01 02.0 +02 52 33 340.3 60.5 0.224 0147330201 106844 38487 discard Abell 1835 14 01 34.6 -11 07 35 329.2 48.1 0.472 0109910101 100219 42837 ok A 1837 14 04 16.6 -34 02 05 319.7 26.4 0.549 0058940301 20049 15753 ok 1401-33 14 08 51.9 -07 52 38 333.8 50.2 0.273 0151590101 32318 10405 ok pks1406-076 14 08 51.9 -07 52 43 333.8 50.2 0.273 0151590201 14299 4257 flared pks1406-076 14 14 52.4 -00 23 45 342.4 55.9 0.322 0145480101 23567 12539 ok Abell 1882 14 15 41.3 +11 29 27 358.6 64.7 0.180 0112251301 29642 25672 ok H1413+117 14 15 45.9 +11 29 28 358.7 64.7 0.180 0112250301 26643 21461 ok H1413+117 14 29 06.5 +01 17 01 349.2 55.1 0.285 0102040501 17951 2069 flared MARK 1383 14 37 33.7 -15 00 31 337.5 40.6 0.778 0081341401 21892 14225 ok IRAS14348-14 14 49 23.7 -10 10 23 344.4 42.9 0.788 0149620201 72555 17408 ok SN 1995N 14 49 23.4 +08 59 53 5.4 56.7 0.202 0148520101 32854 20727 ok Daddi field 14 49 23.4 +08 59 52 5.4 56.7 0.202 0148520301 32664 26914 ok Daddi field 14 49 28.7 +08 59 47 5.4 56.7 0.202 0057560301 43147 34519 ok ERO field 14 52 43.7 +16 45 00 18.9 60.1 0.184 0091140201 32749 26980 ok A1983 14 53 53.2 +03 32 34 359.4 52.4 0.362 0150350101 47159 19737 ok NGC5775 14 54 36.3 +18 38 53 22.8 60.4 0.239 0145020101 42037 27935 ok A 1991 14 58 21.8 -31 40 17 332.2 23.8 0.958 0067750101 53559 20664 ok Cen X-4 15 06 29.5 +01 36 24 0.4 48.7 0.425 0021540101 30178 13115 ok NGC 5846 15 10 51.6 +05 44 22 6.4 50.5 0.307 0111270201 22222 10337 discard A2029 15 16 14.6 +00 05 46 1.1 45.9 0.459 0201902001 28413 23397 ok RXCJ1516.3+0005 15 16 29.4 -00 57 06 0.0 45.2 0.548 0201902101 30416 25126 ok RXCJ1516.5-0056 15 16 35.5 +00 14 41 1.3 45.9 0.459 0103860601 13646 10581 ok CGCG 21-63 15 16 43.9 +07 01 12 9.4 50.1 0.290 0109920101 31597 24994 ok A 2052 241

15 17 00.9 -16 08 25 346.1 34.1 0.840 0164570401 53218 33847 ok GRB040827 15 21 49.6 +07 41 18 11.3 49.4 0.315 0109930101 55413 32636 ok MKW 3s 15 32 22.7 -08 32 07 356.2 37.1 0.868 0100240701 29906 14965 ok UZ LIB 15 32 23.0 -08 31 56 356.2 37.1 0.868 0100240801 30204 22437 ok UZ LIB 15 32 29.2 +04 40 42 9.8 45.5 0.400 0091140401 45414 24947 ok MKW9 15 35 02.0 +01 20 39 6.6 43.0 0.466 0112190401 15111 3593 flared MS1532.5+0130 15 40 48.3 -24 43 07 344.9 24.0 1.347 0152630301 13911 5242 ok SXRB 3 15 50 37.0 -03 53 27 4.3 36.8 0.974 0138951501 12412 2336 flared IRAS15480-0344 15 54 34.7 -25 14 54 347.1 21.5 1.224 0142630301 22210 15005 ok 3 Sco 15 56 25.2 -23 38 00 348.6 22.4 1.159 0112380101 48309 33833 ok SCO-CEN PT1 16 01 14.2 +08 44 22 19.8 41.6 0.376 0147580201 13305 3462 flared 1AXGJ160118+08 16 05 02.3 +17 44 50 31.5 44.5 0.336 0147210301 19008 6312 ok Abell 2151 16 05 02.4 +17 44 49 31.5 44.5 0.336 0147210101 29850 4682 flared Abell 2151 16 05 06.4 +17 45 08 31.5 44.5 0.336 0147210201 29598 1204 flared Abell 2151 16 07 13.2 +08 04 58 20.0 40.0 0.395 0067340601 15391 10580 ok Field VI 16 13 06.9 -83 42 17 308.2 -23.0 0.898 0008820401 54463 35628 ok GRB 020321 16 14 34.2 -06 09 14 6.5 30.7 1.184 0112230901 30809 7591 ok A2163 of3 16 15 46.1 -06 09 09 6.7 30.4 1.227 0112230601 16364 8421 ok A2163 16 15 45.9 -06 09 11 6.7 30.4 1.227 0112231501 13263 5502 ok A2163 16 15 46.1 -05 51 16 7.0 30.6 1.184 0112230801 29400 15264 ok A2163 of2 16 15 46.0 -06 27 13 6.4 30.2 1.227 0112231001 29695 21671 ok A2163 of4 16 16 28.6 +12 12 24 26.1 39.8 0.457 0103460801 14448 10925 ok McNaught-Hartley 16 16 47.8 -06 09 12 6.9 30.2 1.227 0112230701 30909 11901 ok A2163 of1 16 17 02.2 +12 24 23 26.4 39.8 0.457 0103460901 15700 11978 ok McNaught-Hartley 16 17 37.0 +12 37 00 26.7 39.7 0.427 0103461001 13100 9580 ok McNaught-Hartley 16 21 34.3 -15 42 30 359.3 23.4 1.499 0153950201 25040 20409 discard Sco X-1 large 16 21 44.6 -02 16 31 11.4 31.4 0.847 0164560801 26462 6683 ok SN 2004dk 16 32 46.2 +05 34 35 21.0 33.2 0.594 0112230301 22964 16543 ok A2204 16 42 16.7 +03 11 04 20.0 30.0 0.535 0067340501 15278 10211 ok Field V 16 52 58.4 +02 24 18 20.7 27.2 0.549 0101640601 19003 7532 ok NGC 6240 16 52 58.9 +02 24 00 20.7 27.2 0.549 0147420401 14115 7003 ok NGC 6240 16 52 58.8 +02 23 53 20.7 27.2 0.549 0101640101 30111 10396 ok NGC 6240 16 52 58.8 +02 24 07 20.7 27.2 0.549 0147420201 31638 4342 flared NGC 6240 16 52 58.8 +02 24 05 20.7 27.2 0.549 0147420301 28077 1663 flared NGC 6240 16 52 58.7 +02 23 56 20.7 27.2 0.549 0147420501 31219 1542 flared NGC 6240 16 54 00.2 +14 17 47 33.1 32.3 0.551 0113070101 24532 15159 ok nps pos1 17 26 05.4 +02 19 41 25.0 20.0 0.842 0067340401 15280 10474 ok Field IV 18 19 44.6 -63 45 31 330.8 -20.7 0.775 0146340601 18216 5958 ok PKS 1814-637 18 19 43.7 -63 45 14 330.8 -20.7 0.775 0146340501 33855 2365 flared PKS 1814-637 18 47 02.3 -78 31 54 315.7 -26.3 1.008 0147920401 10851 261 flared H1846-786 19 21 13.9 -58 40 17 338.1 -26.7 0.495 0101040501 55489 14130 ok ESO 141-G55 19 31 19.4 -72 39 09 322.5 -28.7 0.597 0081341001 23322 13874 ok IRAS19254-72 19 35 13.0 -52 51 43 344.9 -27.7 0.371 0051610101 22624 14980 ok grb980425 242 APPENDIX A. XDCP SURVEY FIELD LIST

19 39 59.3 -30 57 50 8.7 -23.2 0.858 0111300101 26709 20646 ok M 55 20 03 23.7 -32 51 32 8.4 -28.5 0.749 0104860601 30089 14036 ok [HB89] 2000-330 20 07 47.8 -11 09 16 31.3 -21.9 0.675 0044350501 14921 10438 ok IRAS20051-1117 20 07 50.9 -11 08 20 31.3 -21.9 0.675 0044350201 21638 4395 ﬂared IRAS20051-1117 20 10 57.8 -56 48 12 340.8 -33.1 0.459 0105260501 19503 11975 discard A3667 f5 20 11 09.7 -56 38 40 341.0 -33.2 0.459 0105260201 20146 16130 ok A3667 f2 20 11 21.8 -57 25 34 340.1 -33.2 0.333 0042341101 13916 3615 ﬂared RXCJ2011.3-5725 20 11 54.3 -56 54 54 340.7 -33.3 0.459 0105260601 26467 18472 discard A3667 f6 20 12 13.7 -56 37 25 341.1 -33.3 0.459 0105260401 17469 13205 ok A3667 f4 20 13 04.4 -56 53 11 340.7 -33.4 0.459 0105260101 21303 7504 discard A3667 f1 20 13 08.2 -56 43 17 340.9 -33.4 0.459 0105260301 17871 13421 ok A3667 f3 20 13 29.0 -41 47 22 358.7 -32.4 0.498 0081340501 21913 10696 ok IRAS20100-41 20 13 59.7 -87 57 32 305.0 -27.8 1.141 0083210701 10779 5646 ok 1SAXJ2011.9-8758 20 18 18.1 -70 51 15 324.1 -32.5 0.497 0022340101 32858 18375 ok Pavo group 20 40 10.0 -00 52 23 45.2 -24.4 0.703 0111180201 17270 10630 ok AE Aqr 20 46 20.5 -02 48 54 44.2 -26.7 0.586 0112600501 11360 8305 ok Mrk 896 20 48 14.0 -17 49 51 28.7 -33.5 0.478 0201902401 28257 21534 ok RXCJ2048.1-1750 20 54 18.0 -15 55 36 31.5 -34.2 0.436 0083210101 11462 7154 ok 1SAXJ2054.3-1556 20 56 22.3 -04 37 59 43.8 -29.7 0.502 0112190601 17511 12272 ok MS2053.7-0449 20 58 26.1 -42 38 54 358.6 -40.7 0.382 0081340401 21863 12027 discard IRAS20551-42 21 04 10.3 -11 21 39 37.7 -34.5 0.466 0041150101 40565 30058 ok NGC 7009 21 04 39.5 -12 20 03 36.7 -35.0 0.465 0038540301 17217 12791 ok NGC7010 21 14 55.7 +06 08 31 57.0 -28.0 0.663 0150610201 16914 7408 ok PG 2112+059 21 24 43.5 -33 58 34 10.9 -45.4 0.572 0112320601 70315 20215 ok PSR J2124-3358 21 27 37.2 -44 48 41 355.3 -45.9 0.345 0088020201 32370 10164 ok IRAS F21243-4501 21 29 11.5 -15 38 33 35.9 -41.8 0.493 0103060101 23564 15955 ok PKS 2126-158 21 29 34.7 +00 04 53 53.6 -34.4 0.422 0093030201 58916 34048 ok RX J2129.6+0005 21 32 29.9 +10 09 18 63.6 -29.0 0.465 0150470701 37917 12325 ok UGC 11763 21 37 45.1 -14 32 51 38.4 -43.3 0.477 0092850201 59850 28012 ok PKS 2135-147 21 40 14.7 -23 39 31 26.4 -46.9 0.356 0008830101 22769 9500 ok ms2137-23 21 51 55.1 -30 27 42 17.0 -50.7 0.191 0103060401 25028 19943 ok PKS 2149-306 21 52 01.3 -27 31 59 21.6 -50.3 0.236 0062940401 39421 17220 ok HE2149-2745 21 53 36.5 +17 41 56 73.9 -27.8 0.669 0111270101 23263 9766 ok A2390 22 01 49.7 -59 57 41 332.2 -46.3 0.284 0149670101 25437 19965 ok Abell 3827 22 02 01.8 -31 52 13 15.1 -53.0 0.165 0147920601 16858 11831 ok NGC 7172 22 03 10.6 +18 53 03 76.7 -28.5 0.572 0130920101 20217 13922 ok HD 209458 22 04 30.7 -64 44 31 326.2 -44.0 0.291 0143250101 22216 17825 ok HD 209295 22 05 09.3 -01 55 11 58.0 -42.9 0.607 0012440301 35366 24897 ok PB5062 22 09 15.9 -47 10 00 349.5 -52.5 0.194 0111810101 49610 18189 ok NGC7213 22 15 31.8 -17 44 09 39.2 -52.9 0.228 0106660501 11568 6085 ok LBQS 2212-1759 22 15 32.0 -17 44 06 39.2 -52.9 0.228 0106660101 60508 51704 ok LBQS 2212-1759 22 15 31.9 -17 44 09 39.2 -52.9 0.228 0106660201 53769 37107 ok LBQS 2212-1759 22 15 31.3 -17 44 08 39.2 -52.9 0.228 0106660601 110168 82222 ok LBQS 2212-1759 243

22 15 31.9 -17 44 08 39.2 -52.9 0.228 0106660401 35114 12805 ok LBQS 2212-1759 22 17 04.5 -16 38 35 41.1 -52.8 0.261 0148060201 12800 2657 flared BPS CS 22892-05 22 17 15.4 +14 15 03 76.0 -34.2 0.493 0103660301 47294 10301 ok Mark 304 22 17 31.1 +00 14 15 63.0 -44.0 0.459 0094310101 68821 58820 ok SSA22 22 17 31.1 +00 14 12 63.0 -44.0 0.459 0094310201 81270 60699 ok SSA22 22 19 18.4 +12 08 05 74.7 -36.1 0.536 0103861201 13014 8526 ok IIZw 177 22 20 45.8 -24 40 57 28.4 -56.1 0.199 0049340201 28359 22176 ok NGC 7252 22 28 29.7 -05 18 46 59.1 -49.6 0.514 0100440101 46681 40544 ok PHL 5200 22 29 44.9 -20 50 16 36.1 -57.1 0.268 0125911001 16692 12301 ok NGC 7293 22 35 45.7 -26 02 55 27.1 -59.7 0.147 0111790101 44663 13638 ok NGC7314 22 36 10.8 +13 44 45 79.9 -37.5 0.474 0153220601 12915 5617 ok PG 2233+134 22 36 55.4 -22 13 04 34.6 -59.1 0.216 0103860201 11864 5675 ok ESO 602- G 031 22 37 00.5 -15 16 18 46.8 -56.6 0.400 0056021601 25111 15633 ok RX J2237.0-1516 22 39 56.5 -05 51 30 61.1 -52.1 0.402 0149410401 34475 19256 ok BR 2237-0607 22 40 32.9 +03 22 16 71.8 -46.1 0.524 0110960101 42870 12551 ok Q2237+0305 22 41 59.1 -44 04 00 351.9 -59.0 0.182 0153220101 10919 3344 flared RX J2241.8-4405 22 44 19.2 -72 42 47 314.8 -41.3 0.409 0150050101 59915 20302 ok Galactic Halo 22 49 39.9 -27 06 51 25.8 -62.9 0.178 0111970101 13358 10382 ok TY PsA 22 49 50.0 -64 23 08 322.0 -47.9 0.280 0112240101 35013 25781 ok A3921 22 51 48.7 -17 52 16 45.1 -60.9 0.269 0081340901 23365 19561 ok IRAS22491-18 22 57 09.8 -36 27 33 4.6 -64.1 0.115 0135980201 32903 25360 ok IC1459 23 02 28.9 +08 36 41 82.6 -45.5 0.487 0032140201 13275 6836 ok 1saxJ2302.5+0836 23 03 15.9 +08 52 22 83.0 -45.4 0.529 0112170301 24590 8989 ok NGC7469 23 04 45.6 +03 11 34 78.4 -50.2 0.527 0033541001 13307 10070 ok pg 2302+029 23 04 56.1 +12 19 30 86.2 -42.8 0.519 0025541001 13263 6410 ok NGC7479 23 08 22.4 -02 11 19 73.8 -54.9 0.435 0042341201 12596 1651 flared RXCJ2308.3-0211 23 13 58.5 -42 43 26 348.3 -64.8 0.185 0123900101 66664 29998 ok A S 1101 23 13 59.1 -42 43 43 348.3 -64.8 0.185 0147800101 126314 83040 ok Sersic 159-03 23 15 42.4 -59 04 11 323.7 -54.0 0.243 0081340301 23413 8782 ok IRAS23128-59 23 16 10.7 -42 34 57 348.1 -65.2 0.185 0093640701 20760 7748 ok NGC 7552 23 18 15.2 +00 16 14 79.8 -54.6 0.411 0066950301 12208 2387 flared NGC 7589 23 18 16.4 +00 15 50 79.8 -54.6 0.411 0066950401 13324 4524 ok NGC 7589 23 18 22.2 -42 22 05 348.0 -65.6 0.196 0112310201 23362 18843 ok NGC 7582 23 19 43.4 -73 12 43 311.6 -42.3 0.191 0201903201 31029 0 flared RXCJ2319.6-7313 23 19 50.1 -73 12 55 311.6 -42.3 0.191 0201903301 12616 7975 ok RXCJ2319.6-7313 23 25 17.9 -12 06 23 65.3 -64.8 0.250 0147330101 120513 52958 ok Abell 2597 23 25 21.4 -12 07 20 65.3 -64.8 0.250 0108460201 21073 12680 ok Abell 2597 23 31 49.2 +19 56 29 98.5 -39.1 0.422 0112880301 15772 11809 ok EQ Peg 23 31 52.8 +19 38 10 98.4 -39.4 0.422 0032140701 12311 7348 ok 1saxj2331.9+19 23 31 57.8 +19 44 16 98.5 -39.3 0.422 0147580301 15371 9780 ok 1AXGJ233200+19 23 33 40.6 -15 17 12 62.2 -68.4 0.196 0093550401 25209 19527 ok LEDA 913766 23 36 12.1 +02 08 25 88.1 -55.5 0.493 0112522601 17140 13884 ok NGC 7714 23 36 18.1 +02 10 07 88.2 -55.5 0.493 0112521301 22232 13123 ok NGC 7714 244 APPENDIX A. XDCP SURVEY FIELD LIST

23 37 41.6 +00 16 24 87.0 -57.3 0.385 0042341301 14115 9841 ok RXCJ2337.6+0016 23 37 50.0 -56 24 43 322.1 -57.8 0.204 0083210201 10458 7655 ok 1SAXJ2337.8-56 23 39 54.9 -12 17 38 70.9 -67.6 0.306 0055990301 14965 10380 ok Arp222 23 40 25.9 -11 43 12 72.2 -67.3 0.281 0152200101 53961 45458 ok Abell 2638 23 47 20.2 -02 18 52 88.4 -60.8 0.362 0152860101 33656 24541 ok HCG 97 23 47 22.3 +00 53 03 91.5 -58.0 0.362 0147580401 15369 13043 ok 1AXGJ234725+00 23 51 42.3 -26 04 13 34.0 -76.6 0.166 0148990101 31161 20013 ok Abell 2667 23 54 08.5 -10 23 50 81.3 -68.5 0.292 0108460301 33298 13607 ok Abell 2670 23 56 55.8 -24 24 39 42.3 -77.4 0.173 0125910301 11216 8211 ok HD 224317 23 57 02.9 -34 45 39 356.3 -76.0 0.110 0109950101 30767 18525 ok A 4059 23 57 02.9 -34 45 39 356.3 -76.0 0.110 0109950201 25967 19251 discard A 4059 23 57 27.0 -30 27 29 14.0 -77.7 0.137 0103861501 16336 9388 ﬂared AM2354-304S Appendix B

List of Technical Terms and Acronyms

2MASS: Two Micron All-Sky Survey

AB AB Magnitudes: Magnitude system which uses a fiducial source of constant flux Sν = 2.89 10−21 erg cm−2Hz−1 as reference standard (Schneider, 2006a). This value is chosen× in a way that the V-band magnitudes are equivalent to the Vega system AB Vega mV = mV . Redder filter bands have a positive offset to the Vega system (fainter magnitudes), bluer bands have negative offsets (brighter magnitudes).

ACS: Advanced Camera for Surveys on board the Hubble Space Telescope

ACT: Atacama Cosmology Telescope

AGB: Asymptotic Giant Branch

AGN: Active Galactic Nuclei

APEX: Atacama Pathﬁnder EXperiment

Baﬄe: Light shield placed in the optical path to block stray light and reduce the thermal background radiation.

BAOs: Baryon Acoustic Oscillations

BCG: Brightest Cluster Galaxy

Bias: An oﬀset voltage applied to all pixels in an array detector.

CA: Calar Alto Observatory in Southern Spain

CAL: Calibration Access Layer

CCD: Charge-Coupled Device

245 246 APPENDIX B. LIST OF TECHNICAL TERMS AND ACRONYMS

CDM: Cold Dark Matter

CFHT: Canada-France-Hawaii Telescope

Chandra: American-led X-ray observatory launched in 1999 with an on-axis spatial resolution of 0.5′′.

CMB: Cosmic Microwave Background

COSMOS: Cosmic Evolution Survey

Cosmic: Cosmic ray event in an image. Signal in isolated pixels caused by a charged cosmic particle hit.

CTIO: Cerro Tololo Inter-American Observatory

CXB: Cosmic X-ray Background

DDT: Director’s Discretionary Time

DEC: Declination

DES: Dark Energy Survey

Descriptor: Data buﬀer with a name that contains information about an associated image.

DETF: Dark Energy Task Force

DET ML: Detection Maximum Likelihood

Dithering: Common observation technique in infrared astronomy. Multiple images of an object are taken with slight telescope oﬀsets in between the images. By overlaying several images and using a median process, the local sky for the data reduction can be extracted from the science images.

ECF: Energy Conversion Factor

EMSS: Einstein Medium Sensitivity Survey

EMMI: The ESO Multi Mode Instrument at the NTT telescope in La Silla.

EoS: Equation-of-State relating density and pressure

EPIC: European Photon Imaging Cameras on XMM-Newton eROSITA: extended R¨ontgen Survey with an Imaging Telescope Array

ESA: European Space Agency 247

ESO: European Southern Observatory, with its headquarters in Garching, Germany.

EXT ML: Extent Maximum Likelihood

FITS: Flexible Image Transport System. Standard format for astronomical images.

Flatﬁeld: Image of a uniformly-illuminated area used for the correction of sensitivity variations across the detector.

FORS 2: The FOcal Reducer and low dispersion Spectrograph for the Very Large Tele- scope.

FoV: Field-of-View. Sky area that can be covered with one image of a particular instrument.

FP: False Positive

FPN: Fixed Pattern Noise. Pixel-to-pixel sensitivity variations that are corrected by dividing through a normalized ﬂatﬁeld.

Fringing: Interference patterns produced in CCD layers by strong monochromatic atmospheric emission lines at long optical wavelengths.

FWHM: Full-Width Half-Maximum, a measure of the width of an object in an image. The FWHM is a well-defined number obtained by fitting a Gaussian curve of the form f(x)=1/(σ√2π) exp( x2 ) to the intensity profile of an object. The standard × − 2σ2 deviation σ and the full width at half the peak intensity of the profile only differ by a constant factor of σ =0.424 FWHM (Stöcker, 1998). · Gain: Conversion factor from detected electrons to digital counts.

GDDS: Gemini Deep Deep Survey

GISSEL: Galaxy Isochrone Synthesis Spectral Evolution Library

GMOS: Gemini Multi-Object Spectrograph at the Gemini South 8 m telescope.

GNIRS: Gemini NIR Spectrograph at the Gemini South 8 m telescope.

GR: General Relativity

GROND: Gamma-Ray Burst Optical Near-IR Detector

GTI: Good Time Intervals

HEW: Half Energy Width

HgCdTe: Mercury-cadmium-telluride 248 APPENDIX B. LIST OF TECHNICAL TERMS AND ACRONYMS

HIFLUGCS: HIghest X-ray FLUx Galaxy Cluster Sample

HIROCS: Heidelberg InfraRed / Optical Cluster Survey

HR: Hardness Ratio

HR: Hertzsprung-Russell diagram

HST: Hubble Space Telescope

ICM: Intracluster Medium

Image Catalog: List of image names that can be used within the MIDAS environment to apply a set of commands to all images in the list.

IMF: Stellar Initial Mass Function

IRAC: InfraRed Array Camera on the Spitzer observatory

IRAF: Image Reduction and Analysis Facility. Multi-purpose astronomical software for image reduction and analysis.

I/O: Input/Output

IR: InfraRed

ISAAC: Infrared Spectrometer And Array Camera at the VLT

Keyword: Variable of a speciﬁed data type within the MIDAS environment.

LF: Luminosity Function

LH: Lockman Hole

LM: Limiting Magnitude. Apparent magnitude of objects with a signal-to-noise ratio of 5.

LSS: Large-Scale Structure

M67: Open star cluster in the constellation Cancer at α=8h51m and δ =11.◦82.

Median: Central value in an ordered sequence of numbers.

MEKAL: Plasma emission code from MEwe, KAastra, and Liedahl

MH: Mexican Hat wavelet

MIDAS: Munich Image Data Analysis System. Astronomical software package developed and maintained by the European Southern Observatory.

ML: Maximum Likelihood 249

MOS: Metal Oxid Semi-conductor instrument on XMM-Newton

NED: NASA Extragalactic Data Base

NEP: North Ecliptic Pole survey

NICMOS: Near Infrared Camera and Multi-Object Spectrometer on the Hubble Space Telescope

NIR: Near-InfraRed. In this thesis, usually used for the spectral region 1–2.5 µm, unless otherwise noted.

NORAS: NOrthern ROSAT All-Sky survey

NTT: The 3.5 m New Technology Telescope at La Silla Observatory in Chile.

OBSID: Unique identiﬁcation number for each XMM observation.

ODF: Observation Data File

OMEGA2000: The NIR wide-ﬁeld imager at the 3.5 m Calar Alto telescope.

OoT: Out-of-Time events

PanSTARRS: Panoramic Survey Telescope & Rapid Response System

PI: Principal Investigator

Pipeline: Automated data reduction software requiring only minimal user interaction.

Pixel Coordinates: Coordinate system based on the pixel row and column numbers on the detector.

PN: Imaging camera on XMM-Newton

Pointing: One principal telescope position. The sum of all images with a speciﬁc target in the ﬁeld of view.

PSF: Point-Spread Function. Intensity distribution of a point-like light source in an astronomical image.

PSPC: Position Sensitive Proportional Counter, main instrument of ROSAT.

QE: Quantum Eﬃciency.

QSO: Quasi-Stellar Object 250 APPENDIX B. LIST OF TECHNICAL TERMS AND ACRONYMS

Seeing: “Blurring” of point-like astronomical objects caused by atmospheric turbulence. Without adaptive optics, the seeing limits the highest possible resolution in ground based observations. Typical seeing values for the Calar Alto observatory are somewhat better than 1 arcsec.

RA: Right Ascension

RASS: ROSAT All-Sky Survey

RCS: Red-Sequence Cluster Survey

RDCS: ROSAT Deep Cluster Survey

REFLEX: ROSAT-ESO Flux-Limited X-ray cluster survey

RGS: Reﬂecting Grating Spectrometer on XMM-Newton

ROSAT: ROentgen SATellite, German X-ray survey mission from 1990 to 1999.

RWM: Robertson-Walker-Metric

SAM: Semi-Analytic Model

SAS: XMM-Newton Science Analysis Software

SDSS: Sloan Digital Sky Survey

SED: Spectral Energy Distribution

SEP: South Ecliptic Pole

SFH: Star Formation History

SFR: Star Formation Rate

SGP: South Galactic Pole

SMF: Spectral Matched Filter Scheme

SNR: Signal-to-Noise-Ratio.

SPIE: Society of Photo-Optical Instrumentation Engineers

Spitzer: Mid-infrared space telescope launched in 2003.

SPT: South Pole Telescope

SRT: Special Relativity Theory 251

SSP: Simple Stellar Population models. Galaxy evolution models assuming a single formation redshift of the stellar population and only passive evolution afterwards.

STD: Standard Scheme

SZE: Sunyaev-Zeldovich Eﬀect

UT: Universal time.

VADER: Very Ambitious Dark Energy Research mission

Vega Magnitudes: Magnitude system calibrated to the A0 star Vega for which all filter bands have by definition a magnitude of 0. Since Vega’s spectrum has decreasing flux towards the red optical and NIR filters, these bands have magnitude values which are smaller (i.e. brighter) than the corresponding AB magnitudes. The Vega system is still the standard photometric system for NIR observations.

Vignetting: Obscuration of parts of the primary mirror as seen by a detector pixel.

VISTA: Visible and Infrared Survey Telescope for Astronomy in Chile

VLT: Very Large Telescope in Chile

VST: VLT Survey Telescope

WARPS: Wide Angle ROSAT Pointed Survey

WMAP: Wilkinson Microwave Anisotropy Probe

World Coordinates: Coordinate system that is based on absolute positions on the ce- lestial sphere.

XCS: XMM Cluster Survey

XDCP: XMM-Newton Distant Cluster Project

XLF: X-ray Luminosity Function

XMM-LSS: XMM Large-Scale Structure survey

XMM-Newton: X-ray Spectroscopy Multi-Mirror Mission

XSA: XMM-Newton Science Archive

Bibliography

Abell, G. O.: 1958, Astrophys. J., Suppl. Ser. 3, 211 Abraham, R. G., Glazebrook, K., McCarthy, P. J., Crampton, D., Murowinski, R., Jørgensen, I., Roth, K., Hook, I. M., Savaglio, S., Chen, H.-W., Marzke, R. O., and Carlberg, R. G.: 2004, Astron. J. 127, 2455 Abraham, R. G., Smecker-Hane, T. A., Hutchings, J. B., Carlberg, R. G., Yee, H. K. C., Ellingson, E., Morris, S., Oke, J. B., and Rigler, M.: 1996, Astrophys. J. 471, 694 Allen, S. W., Schmidt, R. W., Ebeling, H., Fabian, A. C., and van Speybroeck, L.: 2004, MNRAS 353, 457 Allen, S. W., Schmidt, R. W., Fabian, A. C., and Ebeling, H.: 2003, MNRAS 342, 287 Andreon, S.: 1998, Astronomy & Astrophysics 336, 98 Andreon, S., Valtchanov, I., Jones, L. R., Altieri, B., Bremer, M., Willis, J., Pierre, M., and Quintana, H.: 2005, MNRAS 359, 1250 Angulo, R., Baugh, C. M., Frenk, C. S., Bower, R. G., Jenkins, A., and Morris, S. L.: 2005, MNRAS 362, L25 Aragon-Salamanca, A., Baugh, C. M., and Kauffmann, G.: 1998, MNRAS 297, 427 Arnaud, M. and Evrard, A. E.: 1999, MNRAS 305, 631 Arnaud, M., Pointecouteau, E., and Pratt, G. W.: 2005, Astronomy & Astrophysics 441, 893 Arnaud, M., Pointecouteau, E., and Pratt, G. W.: 2007, ArXiv Astrophysics e-prints, arXiv:astro-ph/0709.1561 Böhringer, H., Mullis, C. R., Rosati, P., Lamer, G., Fassbender, R., Schwope, A., and Schuecker, P.: 2005, ESO Messenger 120, 33 Böhringer, H., Schuecker, P., Guzzo, L., Collins, C. A., Voges, W., Cruddace, R. G., Ortiz- Gil, A., Chincarini, G., De Grandi, S., Edge, A. C., MacGillivray, H. T., Neumann, D. M., Schindler, S., and Shaver, P.: 2004, Astronomy & Astrophysics 425, 367 Bailer-Jones, C., Bizenberger, P., and Storz, C.: 2000, in M. Iye and A. F. Moorwood (eds.), Optical and IR Telescope Instrumentation and Detectors, Vol. 4008, SPIE - The International Society for Optical Engineering, Bellingham, Washington (USA) Barcons, X., Carrera, F. J., Watson, M. G., McMahon, R. G., Aschenbach, B., Freyberg, M. J., Page, K., Page, M. J., Roberts, T. P., Turner, M. J. L., Barret, D., Brunner, H., Ceballos, M. T., Della Ceca, R., Guillout, P., Hasinger, G., Maccacaro, T., Ma- teos, S., Motch, C., Negueruela, I., Osborne, J. P., Pérez-Fournon, I., Schwope, A., Severgnini, P., Szokoly, G. P., Webb, N. A., Wheatley, P. J., and Worrall, D. M.: 2002,

253 254 BIBLIOGRAPHY

Astronomy & Astrophysics 382, 522 Baum, W. A.: 1959, Publ. Astron. Soc. Pac. 71, 106 Bertin, E. and Arnouts, S.: 1996, Astronomy & Astrophysics, Suppl. Ser. 117, 393 Bhavsar, S. P. and Barrow, J. D.: 1985, MNRAS 213, 857 Biviano, A.: 2000, in Constructing the Universe with Clusters of Galaxies Böhringer, H.: 2003, in R. E. Schielicke (ed.), Reviews in Modern Astronomy, Vol. 16 of Reviews in Modern Astronomy, p. 275 Böhringer, H., Collins, C. A., Guzzo, L., Schuecker, P., Voges, W., Neumann, D. M., Schindler, S., Chincarini, G., De Grandi, S., Cruddace, R. G., Edge, A. C., Reiprich, T. H., and Shaver, P.: 2002a, Astrophys. J. 566, 93 Böhringer, H., Matsushita, K., Churazov, E., Ikebe, Y., and Chen, Y.: 2002b, Astron- omy & Astrophysics 382, 804 Böhringer, H., Pratt, G. W., Finoguenov, A., and Schuecker, P. (eds.): 2007a, Heating versus Cooling in Galaxies and Clusters of Galaxies, Proceedings of the Joint Astronomy Conference Held in Garching, Germany, 6-11 August 2006 Böhringer, H., Schuecker, P., Guzzo, L., Collins, C. A., Voges, W., Schindler, S., Neumann, D. M., Cruddace, R. G., De Grandi, S., Chincarini, G., Edge, A. C., MacGillivray, H. T., and Shaver, P.: 2001, Astronomy & Astrophysics 369, 826 Böhringer, H., Schuecker, P., Pratt, G. W., Arnaud, M., Ponman, T. J., Croston, J. H., Borgani, S., Bower, R. G., Briel, U. G., Collins, C. A., Donahue, M., Forman, W. R., Finoguenov, A., Geller, M. J., Guzzo, L., Henry, J. P., Kneissl, R., Mohr, J. J., Matsushita, K., Mullis, C. R., Ohashi, T., Pedersen, K., Pierini, D., Quintana, H., Raychaudhury, S., Reiprich, T. H., Romer, A. K., Rosati, P., Sabirli, K., Temple, R. F., Viana, P. T. P., Vikhlinin, A., Voit, G. M., and Zhang, Y.-Y.: 2007b, Astron- omy & Astrophysics 469, 363 Böhringer, H., Voges, W., Huchra, J. P., McLean, B., Giacconi, R., Rosati, P., Burg, R., Mader, J., Schuecker, P., Simi¸c, D., Komossa, S., Reiprich, T. H., Retzlaff, J., and Trümper, J.: 2000, Astrophys. J., Suppl. Ser. 129, 435 Bond, J. R., Cole, S., Efstathiou, G., and Kaiser, N.: 1991, Astrophys. J. 379, 440 Borgani, S.: 2006, ArXiv Astrophysics e-prints, astro-ph/0605575 Borgani, S., Rosati, P., della Ceca, R., Tozzi, P., and Norman, C.: 2000, in M. Plionis and I. Georgantopoulos (eds.), Large Scale Structure in the X-ray Universe, Proceedings of the 20-22 September 1999 Workshop, Santorini, Greece, eds. Plionis, M. & Georgan- topoulos, I., Atlantisciences, Paris, France, p. 167 Bower, R. G.: 1991, MNRAS 248, 332 Bower, R. G. and Balogh, M. L.: 2004, in J. S. Mulchaey, A. Dressler, and A. Oemler (eds.), Clusters of Galaxies: Probes of Cosmological Structure and Galaxy Evolution, p. 325 Bower, R. G., Lucey, J. R., and Ellis, R. S.: 1992a, MNRAS 254, 601 Bower, R. G., Lucey, J. R., and Ellis, R. S.: 1992b, MNRAS 254, 589 Bremer, M. N., Valtchanov, I., Willis, J., Altieri, B., Andreon, S., Duc, P. A., Fang, F., Jean, C., Lonsdale, C., Pacaud, F., Pierre, M., Shupe, D. L., Surace, J. A., and Waddington, I.: 2006, MNRAS 371, 1427 BIBLIOGRAPHY 255

Bruzual, G. and Charlot, S.: 2003, MNRAS 344, 1000 Burenin, R. A., Vikhlinin, A., Hornstrup, A., Ebeling, H., Quintana, H., and Mescheryakov, A.: 2007, Astrophys. J., Suppl. Ser. 172, 561 Burke, D. J., Collins, C. A., and Mann, R. G.: 2000, Astrophys. J., Lett. 532, L105 Butcher, H. and Oemler, Jr., A.: 1978, Astrophys. J. 219, 18 Calabretta, M. R. and Greisen, E. W.: 2002, Astronomy & Astrophysics 395, 1077 Caldwell, R. R., Kamionkowski, M., and Weinberg, N. N.: 2003, Physical Review Letters 91, 7 Cassata, P., Guzzo, L., Franceschini, A., Scoville, N., Capak, P., Ellis, R. S., Koeke- moer, A., McCracken, H. J., Mobasher, B., Renzini, A., Ricciardelli, E., Scodeggio, M., Taniguchi, Y., and Thompson, D.: 2007, Astrophys. J., Suppl. Ser. 172, 270 Cavaliere, A. and Fusco-Femiano, R.: 1976, Astronomy & Astrophysics 49, 137 Chandrasekhar, S.: 1943, Astrophys. J. 97, 255 Chen, Y., Reiprich, T. H., Böhringer, H., Ikebe, Y., and Zhang, Y.-Y.: 2007, Astron- omy & Astrophysics 466, 805 Churazov, E., Brüggen, M., Kaiser, C. R., Böhringer, H., and Forman, W.: 2001, Astro- phys. J. 554, 261 Cimatti, A., Daddi, E., Renzini, A., Cassata, P., Vanzella, E., Pozzetti, L., Cristiani, S., Fontana, A., Rodighiero, G., Mignoli, M., and Zamorani, G.: 2004, Nature 430, 184 Clowe, D., Bradaˇc, M., Gonzalez, A. H., Markevitch, M., Randall, S. W., Jones, C., and Zaritsky, D.: 2006, Astrophys. J., Lett. 648, L109 Coles, P. and Lucchin, F.: 2002, Cosmology: The Origin and Evolution of Cosmic Struc- ture, Second Edition, Cosmology: The Origin and Evolution of Cosmic Structure, Sec- ond Edition, by Peter Coles, Francesco Lucchin, pp. 512. ISBN 0-471-48909-3. Wiley- VCH , July 2002. Collins, C. A. and Mann, R. G.: 1998, MNRAS 297, 128 Condon, J. J., Cotton, W. D., Greisen, E. W., Yin, Q. F., Perley, R. A., Taylor, G. B., and Broderick, J. J.: 1998, Astron. J. 115, 1693 Connolly, A. J., Szalay, A. S., Koo, D., Romer, A. K., Holden, B., Nichol, R. C., and Miyaji, T.: 1996, Astrophys. J., Lett. 473, L67 Cooper, M. C., Newman, J. A., Coil, A. L., Croton, D. J., Gerke, B. F., Yan, R., Davis, M., Faber, S. M., Guhathakurta, P., Koo, D. C., Weiner, B. J., and Willmer, C. N. A.: 2007, MNRAS 376, 1445 Cowie, L. L., Songaila, A., Hu, E. M., and Cohen, J. G.: 1996, Astron. J. 112, 839 Cox, A. N. (ed.): 2000, Allen’s Astrophysical Quantities, New York: AIP Press; Springer Cruddace, R. G., Hasinger, G. R., and Schmitt, J. H.: 1988, in F. Murtagh and A. Heck (eds.), Astronomy from Large Databases, pp 177–182 Cucciati, O., Iovino, A., Marinoni, C., Ilbert, O., Bardelli, S., Franzetti, P., Le Fevre, O., Pollo, A., Zamorani, G., Cappi, A., Guzzo, L., McCracken, H. J., Meneux, B., Scaramella, R., Scodeggio, M., Tresse, L., Zucca, E., Bottini, D., Garilli, B., Le Brun, V., Maccagni, D., Picat, J. P., Vettolani, G., Zanichelli, A., Adami, C., Arnaboldi, M., Arnouts, S., Bolzonella, M., Charlot, S., Ciliegi, P., Contini, T., Foucaud, S., Gavignaud, I., Marano, B., Mazure, A., Merighi, R., Paltani, S., Pello’, R., Pozzetti, 256 BIBLIOGRAPHY

L., Radovich, M., Bondi, M., Bongiorno, A., Busarello, G., de la Torre, S., Gregorini, L., Lamareille, F., Mathez, G., Mellier, Y., Merluzzi, P., Ripepi, V., Rizzo, D., Temporin, G., and Vergani, D.: 2006, ArXiv Astrophysics e-prints, astro-ph/0612120 Cutri, R. M., Skrutskie, M. F., van Dyk, S., Beichman, C. A., Carpenter, J. M., Chester, T., Cambresy, L., Evans, T., Fowler, J., Gizis, J., Howard, E., Huchra, J., Jarrett, T., Kopan, E. L., Kirkpatrick, J. D., Light, R. M., Marsh, K. A., McCallon, H., Schneider, S., Stiening, R., Sykes, M., Weinberg, M., Wheaton, W. A., Wheelock, S., and Zacarias, N.: 2003, 2MASS All Sky Catalog of point sources., The IRSA 2MASS All-Sky Point Source Catalog, NASA/IPAC Infrared Science Archive Danese, L., de Zotti, G., and di Tullio, G.: 1980, Astronomy & Astrophysics 82, 322 Davis, T. M. and Lineweaver, C. H.: 2004, Publications of the Astronomical Society of Australia 21, 97 De Lucia, G. and Blaizot, J.: 2007, MNRAS 375, 2 De Lucia, G., Poggianti, B. M., Aragón-Salamanca, A., Clowe, D., Halliday, C., Jablonka, P., Milvang-Jensen, B., Pelló, R., Poirier, S., Rudnick, G., Saglia, R., Simard, L., and White, S. D. M.: 2004, Astrophys. J., Lett. 610, L77 De Lucia, G., Poggianti, B. M., Aragón-Salamanca, A., White, S. D. M., Zaritsky, D., Clowe, D., Halliday, C., Jablonka, P., von der Linden, A., Milvang-Jensen, B., Pelló, R., Rudnick, G., Saglia, R. P., and Simard, L.: 2007, MNRAS 374, 809 De Lucia, G., Springel, V., White, S. D. M., Croton, D., and Kauffmann, G.: 2006, MNRAS 366, 499 Dickey, J. M. and Lockman, F. J.: 1990, Annual Rev. of Astronomy and Astrophysics 28, 215 Dressler, A.: 1980, Astrophys. J. 236, 351 Dressler, A., Oemler, A. J., Couch, W. J., Smail, I., Ellis, R. S., Barger, A., Butcher, H., Poggianti, B. M., and Sharples, R. M.: 1997, Astrophys. J. 490, 577 Dunlop, J., Peacock, J., Spinrad, H., Dey, A., Jimenez, R., Stern, D., and Windhorst, R.: 1996, Nature 381, 581 Eggen, O. J., Lynden-Bell, D., and Sandage, A. R.: 1962, Astrophys. J. 136, 748 Ehle, M., e. a.: 2007, XMM-Newton User’s Handbook, European Space Agency Eisenstein, D. J. and Hu, W.: 1998, Astrophys. J. 496, 605 Eisenstein, D. J., Zehavi, I., Hogg, D. W., Scoccimarro, R., Blanton, M. R., Nichol, R. C., Scranton, R., Seo, H.-J., Tegmark, M., Zheng, Z., Anderson, S. F., Annis, J., Bahcall, N., Brinkmann, J., Burles, S., Castander, F. J., Connolly, A., Csabai, I., Doi, M., Fukugita, M., Frieman, J. A., Glazebrook, K., Gunn, J. E., Hendry, J. S., Hennessy, G., Ivezić, Z., Kent, S., Knapp, G. R., Lin, H., Loh, Y.-S., Lupton, R. H., Margon, B., McKay, T. A., Meiksin, A., Munn, J. A., Pope, A., Richmond, M. W., Schlegel, D., Schneider, D. P., Shimasaku, K., Stoughton, C., Strauss, M. A., SubbaRao, M., Szalay, A. S., Szapudi, I., Tucker, D. L., Yanny, B., and York, D. G.: 2005, Astrophys. J. 633, 560 Ettori, S., Tozzi, P., Borgani, S., and Rosati, P.: 2004, Astronomy & Astrophysics 417, 13 Faber, S. M., Trager, S. C., Gonzalez, J. J., and Worthey, G.: 1995, in P. C. van der Kruit and G. Gilmore (eds.), Stellar Populations, Vol. 164 of IAU Symposium, p. 249 BIBLIOGRAPHY 257

Fabian, A. C. and Nulsen, P. E. J.: 1977, MNRAS 180, 479 Fabian, A. C., Sanders, J. S., Allen, S. W., Crawford, C. S., Iwasawa, K., Johnstone, R. M., Schmidt, R. W., and Taylor, G. B.: 2003, MNRAS 344, L43 Fassbender, R.: 2003, Master’s thesis, University of Heidelberg Fassbender, R., Böhringer, H., Lamer, G., Mullis, C. R., Rosati, P., Schope, A., Kohnert, J., and Santos, J. S.: submitted, Astronomy & Astrophysics Fassbender, R., Stegmaier, J., Weijmans, A.-M., Köstner, S., Kruselburger, A., Diethart, C., Fertl, P., Valiante, E., Hayes, M., Schuecker, P., and Hasinger, G.: 2006, in M. J. L. Turner and G. Hasinger (eds.), Space Telescopes and Instrumentation II: Ultraviolet to Gamma Ray. Edited by Turner, Martin J. L.; Hasinger, Günther. Proceedings of the SPIE, Volume 6266, pp. 626632 (2006)., Vol. 6266 of Presented at the Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Felten, J. E., Gould, R. J., Stein, W. A., and Woolf, N. J.: 1966, Astrophys. J. 146, 955 Finoguenov, A., Guzzo, L., Hasinger, G., Scoville, N. Z., Aussel, H., Böhringer, H., Brusa, M., Capak, P., Cappelluti, N., Comastri, A., Giodini, S., Griffiths, R. E., Impey, C., Koekemoer, A. M., Kneib, J.-P., Leauthaud, A., Le Fèvre, O., Lilly, S., Mainieri, V., Massey, R., McCracken, H. J., Mobasher, B., Murayama, T., Peacock, J. A., Sakelliou, I., Schinnerer, E., Silverman, J. D., Smolˇcić, V., Taniguchi, Y., Tasca, L., Taylor, J. E., Trump, J. R., and Zamorani, G.: 2007, Astrophys. J., Suppl. Ser. 172, 182 Fioc, M. and Rocca-Volmerange, B.: 1997, in ASSL Vol. 210: The Impact of Large Scale Near-IR Sky Surveys, p. 257 Freedman, W. L., Madore, B. F., Gibson, B. K., Ferrarese, L., Kelson, D. D., Sakai, S., Mould, J. R., Kennicutt, Jr., R. C., Ford, H. C., Graham, J. A., Huchra, J. P., Hughes, S. M. G., Illingworth, G. D., Macri, L. M., and Stetson, P. B.: 2001, Astrophys. J. 553, 47 Gabasch, A.: 2004, Ph.D. thesis, Ludwig-Maximilian University Munich Giacconi, R., Rosati, P., Tozzi, P., Nonino, M., Hasinger, G., Norman, C., Bergeron, J., Borgani, S., Gilli, R., Gilmozzi, R., and Zheng, W.: 2001, Astrophys. J. 551, 624 Gioia, I. M., Henry, J. P., Maccacaro, T., Morris, S. L., Stocke, J. T., and Wolter, A.: 1990, Astrophys. J., Lett. 356, L35 Gladders, M. D. and Yee, H. K. C.: 2005, Astrophys. J., Suppl. Ser. 157, 1 Gómez, P. L., Nichol, R. C., Miller, C. J., Balogh, M. L., Goto, T., Zabludoff, A. I., Romer, A. K., Bernardi, M., Sheth, R., Hopkins, A. M., Castander, F. J., Connolly, A. J., Schneider, D. P., Brinkmann, J., Lamb, D. Q., SubbaRao, M., and York, D. G.: 2003, Astrophys. J. 584, 210 Gonzalez, A. H., Zaritsky, D., Dalcanton, J. J., and Nelson, A.: 2001, Astrophys. J., Suppl. Ser. 137, 117 Gott, J. R. I. and Gunn, J. E.: 1971, Astrophys. J., Lett. 169, L13 Gunn, J. E. and Gott, J. R. I.: 1972, Astrophys. J. 176, 1 Haas, A.: 2002, HAWAII-2: 2048x2048 Pixel Focal Plane Array, Rockwell Scientific Harris, W. E.: 1990, Publ. Astron. Soc. Pac. 102, 949 Hashimoto, Y., Hasinger, G., Arnaud, M., Rosati, P., and Miyaji, T.: 2002, Astron- omy & Astrophysics 381, 841 258 BIBLIOGRAPHY

Hashimoto, Y., Henry, J. P., Hasinger, G., Szokoly, G., and Schmidt, M.: 2005, Astron- omy & Astrophysics 439, 29 Hasinger, G., Cappelluti, N., Brunner, H., Brusa, M., Comastri, A., Elvis, M., Finoguenov, A., Fiore, F., Franceschini, A., Gilli, R., Griffiths, R. E., Lehmann, I., Mainieri, V., Matt, G., Matute, I., Miyaji, T., Molendi, S., Paltani, S., Sanders, D. B., Scoville, N., Tresse, L., Urry, C. M., Vettolani, P., and Zamorani, G.: 2007, Astrophys. J., Suppl. Ser. 172, 29 Hausman, M. A. and Ostriker, J. P.: 1978, Astrophys. J. 224, 320 Henry, J. P., Gioia, I. M., Maccacaro, T., Morris, S. L., Stocke, J. T., and Wolter, A.: 1992, Astrophys. J. 386, 408 Henry, J. P., Mullis, C. R., Voges, W., Böhringer, H., Briel, U. G., Gioia, I. M., and Huchra, J. P.: 2006, Astrophys. J., Suppl. Ser. 162, 304 Herranz, D., Sanz, J. L., Barreiro, R. B., and Mart´ınez-González, E.: 2002, Astrophys. J. 580, 610 Hu, W.: 2004, ArXiv Astrophysics e-prints, astro-ph/0402060 Hu, W. and Sugiyama, N.: 1996, Astrophys. J. 471, 542 Jehin, E.: 2006, Fors1+2 User Manual, European Southern Observatory Jenkins, A., Frenk, C. S., White, S. D. M., Colberg, J. M., Cole, S., Evrard, A. E., Couchman, H. M. P., and Yoshida, N.: 2001, MNRAS 321, 372 Jones, C. and Forman, W.: 1984, Astrophys. J. 276, 38 Kaiser, N.: 1984, Astrophys. J., Lett. 284, L9 Kaiser, N.: 1986, MNRAS 222, 323 Kajisawa, M., Yamada, T., Tanaka, I., Maihara, T., Iwamuro, F., Terada, H., Goto, M., Motohara, K., Tanabe, H., Taguchi, T., Hata, R., Iye, M., Imanishi, M., Chikada, Y., Yoshida, M., Simpson, C., Sasaki, T., Kosugi, G., Usuda, T., and Sekiguchi, K.: 2000, Publ. Astron. Soc. Jpn. 52, 61 Kauffmann, G.: 1996, MNRAS 281, 487 Kauffmann, G. and Charlot, S.: 1998, ArXiv Astrophysics e-prints, astro-ph/9810031 King, I. R.: 1966, Astron. J. 71, 276 Kitayama, T. and Suto, Y.: 1996, Astrophys. J. 469, 480 Kodama, T. and Arimoto, N.: 1997, Astronomy & Astrophysics 320, 41 Kodama, T., Arimoto, N., Barger, A. J., and Arag’on-Salamanca, A.: 1998, Astron- omy & Astrophysics 334, 99 Kodama, T., Smail, I., Nakata, F., Okamura, S., and Bower, R. G.: 2001, Astrophys. J., Lett. 562, L9 Kurtz, M. J. and Mink, D. J.: 1998, Publ. Astron. Soc. Pac. 110, 934 Lamer, G., Soucail, G., Fassbender, R., and et al.: in preparation, Astron- omy & Astrophysics Lanzetta, K. M., Yahata, N., Pascarelle, S., Chen, H.-W., and Fernández-Soto, A.: 2002, Astrophys. J. 570, 492 Lewis, I., Balogh, M., De Propris, R., Couch, W., Bower, R., Offer, A., Bland-Hawthorn, J., Baldry, I. K., Baugh, C., Bridges, T., Cannon, R., Cole, S., Colless, M., Collins, C., Cross, N., Dalton, G., Driver, S. P., Efstathiou, G., Ellis, R. S., Frenk, C. S., Glazebrook, BIBLIOGRAPHY 259

K., Hawkins, E., Jackson, C., Lahav, O., Lumsden, S., Maddox, S., Madgwick, D., Norberg, P., Peacock, J. A., Percival, W., Peterson, B. A., Sutherland, W., and Taylor, K.: 2002, MNRAS 334, 673 Lin, Y.-T. and Mohr, J. J.: 2004, Astrophys. J. 617, 879 Lin, Y.-T. and Mohr, J. J.: 2007, Astrophys. J., Suppl. Ser. 170, 71 Lynden-Bell, D.: 1967, MNRAS 136, 101 Margoniner, V. E., de Carvalho, R. R., Gal, R. R., and Djorgovski, S. G.: 2001, Astrophys. J., Lett. 548, L143 Matthews, T. A., Morgan, W. W., and Schmidt, M.: 1964, Astrophys. J. 140, 35 Maughan, B. J., Jones, L. R., Ebeling, H., and Scharf, C.: 2006, MNRAS 365, 509 McCarthy, P. J., Yan, H., Abraham, R. G., Mentuch, E., Glazebrook, K., Yan, L., Chen, H.-W., Persson, S. E., Nair, P., Savaglio, S., Crampton, D., Juneau, S., Le Borgne, D., Carlberg, R. G., Marzke, R. O., Jørgensen, I., Roth, K., and Murowinski, R.: 2007, Astrophys. J., Lett. 664, L17 McNamara, B. R., Birzan, L., Rafferty, D. A., Nulsen, P. E. J., Carilli, C., and Wise, M. W.: 2007, ArXiv e-prints, astro-ph/0708.0579 McNamara, B. R., Nulsen, P. E. J., Wise, M. W., Rafferty, D. A., Carilli, C., Sarazin, C. L., and Blanton, E. L.: 2005, Nature 433, 45 Mei, S., Blakeslee, J. P., Stanford, S. A., Holden, B. P., Rosati, P., Strazzullo, V., Homeier, N., Postman, M., Franx, M., Rettura, A., Ford, H., Illingworth, G. D., Ettori, S., Bouwens, R. J., Demarco, R., Martel, A. R., Clampin, M., Hartig, G. F., Eisenhardt, P., Ardila, D. R., Bartko, F., Ben´ıtez, N., Bradley, L. D., Broadhurst, T. J., Brown, R. A., Burrows, C. J., Cheng, E. S., Cross, N. J. G., Feldman, P. D., Golimowski, D. A., Goto, T., Gronwall, C., Infante, L., Kimble, R. A., Krist, J. E., Lesser, M. P., Menanteau, F., Meurer, G. R., Miley, G. K., Motta, V., Sirianni, M., Sparks, W. B., Tran, H. D., Tsvetanov, Z. I., White, R. L., and Zheng, W.: 2006a, Astrophys. J. 639, 81 Mei, S., Holden, B. P., Blakeslee, J. P., Rosati, P., Postman, M., Jee, M. J., Rettura, A., Sirianni, M., Demarco, R., Ford, H. C., Franx, M., Homeier, N., and Illingworth, G. D.: 2006b, Astrophys. J. 644, 759 Miley, G. K., Overzier, R. A., Zirm, A. W., Ford, H. C., Kurk, J., Pentericci, L., Blakeslee, J. P., Franx, M., Illingworth, G. D., Postman, M., Rosati, P., Röttgering, H. J. A., Venemans, B. P., and Helder, E.: 2006, Astrophys. J., Lett. 650, L29 Mitchell, R. J. and Culhane, J. L.: 1977, MNRAS 178, 75P Mullis, C. R., Rosati, P., Lamer, G., Böhringer, H., Schwope, A., Schuecker, P., and Fassbender, R.: 2005, Astrophys. J., Lett. 623, L85 Mullis, C. R., Vikhlinin, A., Henry, J. P., Forman, W., Gioia, I. M., Hornstrup, A., Jones, C., McNamara, B. R., and Quintana, H.: 2004, Astrophys. J. 607, 175 Mushotzky, R. F., Cowie, L. L., Barger, A. J., and Arnaud, K. A.: 2000, Nature 404, 459 Nakata, F., Kajisawa, M., Yamada, T., Kodama, T., Shimasaku, K., Tanaka, I., Doi, M., Furusawa, H., Hamabe, M., Iye, M., Kimura, M., Komiyama, Y., Miyazaki, S., Okamura, S., Ouchi, M., Sasaki, T., Sekiguchi, M., Yagi, M., and Yasuda, N.: 2001, Publ. Astron. Soc. Jpn. 53, 1139 260 BIBLIOGRAPHY

Nakata, F., Kodama, T., Shimasaku, K., Doi, M., Furusawa, H., Hamabe, M., Kimura, M., Komiyama, Y., Miyazaki, S., Okamura, S., Ouchi, M., Sekiguchi, M., Ueda, Y., Yagi, M., and Yasuda, N.: 2005, MNRAS 357, 1357 Navarro, J. F., Frenk, C. S., and White, S. D. M.: 1997, Astrophys. J. 490, 493 Nipoti, C., Stiavelli, M., Ciotti, L., Treu, T., and Rosati, P.: 2003, MNRAS 344, 748 Novak, N.: 2003, Master’s thesis, Technical University Munich Ostriker, J. P. and Tremaine, S. D.: 1975, Astrophys. J., Lett. 202, L113 Pacaud, F., Pierre, M., Adami, C., Altieri, B., Andreon, S., Chiappetti, L., Detal, A., Duc, P. ., Galaz, G., Gueguen, A., Le Fèvre, J. ., Hertling, G., Libbrecht, C., Melin, J. ., Ponman, T. J., Quintana, H., Refregier, A., Sprimont, P. ., Surdej, J., Valtchanov, I., Willis, J. P., Alloin, D., Birkinshaw, M., Bremer, M. N., Garcet, O., Jean, C., Jones, L. R., Le Fèvre, O., Maccagni, D., Mazure, A., Proust, D., Röttgering, H. J. A., and Trinchieri, G.: 2007, ArXiv e-prints, astro-ph/0709.1950 Pacaud, F., Pierre, M., Refregier, A., Gueguen, A., Starck, J.-L., Valtchanov, I., Read, A. M., Altieri, B., Chiappetti, L., Gandhi, P., Garcet, O., Gosset, E., Ponman, T. J., and Surdej, J.: 2006, MNRAS 372, 578 Peacock, J. A.: 1999, Cosmological Physics, Cosmological Physics, by John A. Peacock, pp. 704. ISBN 052141072X. Cambridge, UK: Cambridge University Press, January 1999. Perlman, E. S., Horner, D. J., Jones, L. R., Scharf, C. A., Ebeling, H., Wegner, G., and Malkan, M.: 2002, Astrophys. J., Suppl. Ser. 140, 265 Perlmutter, S., Aldering, G., Goldhaber, G., Knop, R. A., Nugent, P., Castro, P. G., Deustua, S., Fabbro, S., Goobar, A., Groom, D. E., Hook, I. M., Kim, A. G., Kim, M. Y., Lee, J. C., Nunes, N. J., Pain, R., Pennypacker, C. R., Quimby, R., Lidman, C., Ellis, R. S., Irwin, M., McMahon, R. G., Ruiz-Lapuente, P., Walton, N., Schaefer, B., Boyle, B. J., Filippenko, A. V., Matheson, T., Fruchter, A. S., Panagia, N., Newberg, H. J. M., Couch, W. J., and The Supernova Cosmology Project: 1999, Astrophys. J. 517, 565 Peterson, J. R., Paerels, F. B. S., Kaastra, J. S., Arnaud, M., Reiprich, T. H., Fabian, A. C., Mushotzky, R. F., Jernigan, J. G., and Sakelliou, I.: 2001, Astronomy & Astrophysics 365, L104 Pickles, A. J.: 1998, Publ. Astron. Soc. Pac. 110, 863 Pierre, M., Pacaud, F., Duc, P.-A., Willis, J. P., Andreon, S., Valtchanov, I., Altieri, B., Galaz, G., Gueguen, A., Fèvre, J.-P. L., Fèvre, O. L., Ponman, T., Sprimont, P.-G., Surdej, J., Adami, C., Alshino, A., Bremer, M., Chiappetti, L., Detal, A., Garcet, O., Gosset, E., Jean, C., Maccagni, D., Marinoni, C., Mazure, A., Quintana, H., and Read, A.: 2006, MNRAS 372, 591 Pointecouteau, E., Arnaud, M., and Pratt, G. W.: 2005, Astronomy & Astrophysics 435,1 Popesso, P., Böhringer, H., Romaniello, M., and Voges, W.: 2005, Astron- omy & Astrophysics 433, 415 Postman, M., Franx, M., Cross, N. J. G., Holden, B., Ford, H. C., Illingworth, G. D., Goto, T., Demarco, R., Rosati, P., Blakeslee, J. P., Tran, K.-V., Ben´ıtez, N., Clampin, M., Hartig, G. F., Homeier, N., Ardila, D. R., Bartko, F., Bouwens, R. J., Bradley, L. D., Broadhurst, T. J., Brown, R. A., Burrows, C. J., Cheng, E. S., Feldman, P. D., BIBLIOGRAPHY 261

Golimowski, D. A., Gronwall, C., Infante, L., Kimble, R. A., Krist, J. E., Lesser, M. P., Martel, A. R., Mei, S., Menanteau, F., Meurer, G. R., Miley, G. K., Motta, V., Sirianni, M., Sparks, W. B., Tran, H. D., Tsvetanov, Z. I., White, R. L., and Zheng, W.: 2005, Astrophys. J. 623, 721 Postman, M. and Lauer, T. R.: 1995, Astrophys. J. 440, 28 Pratt, G. W. and Arnaud, M.: 2002, Astronomy & Astrophysics 394, 375 Pratt, G. W. and Arnaud, M.: 2003, Astronomy & Astrophysics 408, 1 Pratt, G. W., Böhringer, H., Croston, J. H., Arnaud, M., Borgani, S., Finoguenov, A., and Temple, R. F.: 2007, Astronomy & Astrophysics 461, 71 Predehl, P., Hasinger, G., Böhringer, H., Briel, U., Brunner, H., Churazov, E., Freyberg, M., Friedrich, P., Kendziorra, E., Lutz, D., Meidinger, N., Pavlinsky, M., Pfeffermann, E., Santangelo, A., Schmitt, J., Schuecker, P., Schwope, A., Steinmetz, M., Strüder, L., Sunyaev, R., and Wilms, J.: 2006, in Space Telescopes and Instrumentation II: Ultra- violet to Gamma Ray. Edited by Turner, Martin J. L.; Hasinger, Günther. Proceedings of the SPIE, Volume 6266, pp. 62660P (2006). Press, W. H. and Schechter, P.: 1974, Astrophys. J. 187, 425 Reid, I. N., Brewer, C., Brucato, R. J., McKinley, W. R., Maury, A., Mendenhall, D., Mould, J. R., Mueller, J., Neugebauer, G., Phinney, J., Sargent, W. L. W., Schombert, J., and Thicksten, R.: 1991, Publ. Astron. Soc. Pac. 103, 661 Reiprich, T. H.: 2001, Ph.D. thesis, Ludwig-Maximilian University Munich Reiprich, T. H. and Böhringer, H.: 2002, Astrophys. J. 567, 716 Renzini, A.: 2006, Annual Rev. of Astronomy and Astrophysics 44, 141 Richstone, D. O.: 1976, Astrophys. J. 204, 642 Romer, A. K., Viana, P. T. P., Liddle, A. R., and Mann, R. G.: 2001, Astrophys. J. 547, 594 Rosati, P.: 2001, in ASP Conf. Ser. 234: X-ray Astronomy 2000, p. 363 Rosati, P., Borgani, S., della Ceca, R., Stanford, A., Eisenhardt, P., and Lidman, C.: 2000, in M. Plionis and I. Georgantopoulos (eds.), Large Scale Structure in the X-ray Universe, Proceedings of the 20-22 September 1999 Workshop, Santorini, Greece, eds. Plionis, M. & Georgantopoulos, I., Atlantisciences, Paris, France, p. 13 Rosati, P., Borgani, S., and Norman, C.: 2002, Annual Rev. of Astronomy and Astrophysics 40, 539 Rosati, P., della Ceca, R., Burg, R., Norman, C., and Giacconi, R.: 1995, Astrophys. J., Lett. 445, L11 Rosati, P., della Ceca, R., Norman, C., and Giacconi, R.: 1998, Astrophys. J., Lett. 492, L21 Rosati, P., Stanford, S. A., Eisenhardt, P. R., Elston, R., Spinrad, H., Stern, D., and Dey, A.: 1999, Astron. J. 118, 76 Rosati, P., Tozzi, P., Ettori, S., Mainieri, V., Demarco, R., Stanford, S. A., Lidman, C., Nonino, M., Borgani, S., Della Ceca, R., Eisenhardt, P., Holden, B. P., and Norman, C.: 2004, Astron. J. 127, 230 Röser, H.-J.: 2005, OMEGA2000 User’s Manual, MPI für Astronomie Ruhl, J., Ade, P. A. R., Carlstrom, J. E., Cho, H.-M., Crawford, T., Dobbs, M., Greer, 262 BIBLIOGRAPHY

C. H., Halverson, N. w., Holzapfel, W. L., Lanting, T. M., Lee, A. T., Leitch, E. M., Leong, J., Lu, W., Lueker, M., Mehl, J., Meyer, S. S., Mohr, J. J., Padin, S., Plagge, T., Pryke, C., Runyan, M. C., Schwan, D., Sharp, M. K., Spieler, H., Staniszewski, Z., and Stark, A. A.: 2004, in C. M. Bradford, P. A. R. Ade, J. E. Aguirre, J. J. Bock, M. Dragovan, L. Duband, L. Earle, J. Glenn, H. Matsuhara, B. J. Naylor, H. T. Nguyen, M. Yun, and J. Zmuidzinas (eds.), Millimeter and Submillimeter Detectors for Astronomy II. Edited by Jonas Zmuidzinas, Wayne S. Holland and Stafford Withington Proceedings of the SPIE, Volume 5498, pp. 11-29 (2004)., Vol. 5498 of Presented at the Society of Photo-Optical Instrumentation Engineers (SPIE) Conference, pp 11–29 Sandage, A. and Visvanathan, N.: 1978, Astrophys. J. 223, 707 Santos, J. S., Rosati, P., Tozzi, P., Böhringer, H., and Ettori, S.: submitted, Astron- omy & Astrophysics Sarazin, C. L.: 1986, Reviews of Modern Physics 58, 1 Sasaki, S.: 1996, Publ. Astron. Soc. Jpn. 48, L119 Scharf, C.: 2002, Astrophys. J. 572, 157 Schlegel, D. J., Finkbeiner, D. P., and Davis, M.: 1998, Astrophys. J. 500, 525 Schneider, P.: 2006a, Einführung in die extragalaktische Astronomie und Kosmologie, Peter Schneider. Berlin: Springer Schneider, P.: 2006b, in G. Meylan, P. Jetzer, P. North, P. Schneider, C. S. Kochanek, and J. Wambsganss (eds.), Saas-Fee Advanced Course 33: Gravitational Lensing: Strong, Weak and Micro, pp 269–451 Schombert, J. M.: 1986, Astrophys. J., Suppl. Ser. 60, 603 Schombert, J. M.: 1988, Astrophys. J. 328, 475 Schuecker, P. and Böhringer, H.: 2003, in S. Bowyer and C.-Y. Hwang (eds.), Astronomical Society of the Pacific Conference Series, Vol. 301 of Astronomical Society of the Pacific Conference Series, p. 97 Schuecker, P., Böhringer, H., Guzzo, L., Collins, C. A., Neumann, D. M., Schindler, S., Voges, W., De Grandi, S., Chincarini, G., Cruddace, R., Müller, V., Reiprich, T. H., Retzlaff, J., and Shaver, P.: 2001, Astronomy & Astrophysics 368, 86 Schuecker, P., Caldwell, R. R., Böhringer, H., Collins, C. A., Guzzo, L., and Weinberg, N. N.: 2003, Astronomy & Astrophysics 402, 53 Scoville, N., Aussel, H., Brusa, M., Capak, P., Carollo, C. M., Elvis, M., Giavalisco, M., Guzzo, L., Hasinger, G., Impey, C., Kneib, J.-P., LeFevre, O., Lilly, S. J., Mobasher, B., Renzini, A., Rich, R. M., Sanders, D. B., Schinnerer, E., Schminovich, D., Shopbell, P., Taniguchi, Y., and Tyson, N. D.: 2007, Astrophys. J., Suppl. Ser. 172, 1 Sheth, R. K., Mo, H. J., and Tormen, G.: 2001, MNRAS 323, 1 Silk, J.: 1976, Astrophys. J. 208, 646 Smith, J. A., Tucker, D. L., Kent, S., Richmond, M. W., Fukugita, M., Ichikawa, T., Ichikawa, S.-i., Jorgensen, A. M., Uomoto, A., Gunn, J. E., Hamabe, M., Watanabe, M., Tolea, A., Henden, A., Annis, J., Pier, J. R., McKay, T. A., Brinkmann, J., Chen, B., Holtzman, J., Shimasaku, K., and York, D. G.: 2002, Astron. J. 123, 2121 Spergel, D. N., Bean, R., Doré, O., Nolta, M. R., Bennett, C. L., Dunkley, J., Hinshaw, G., Jarosik, N., Komatsu, E., Page, L., Peiris, H. V., Verde, L., Halpern, M., Hill, BIBLIOGRAPHY 263

R. S., Kogut, A., Limon, M., Meyer, S. S., Odegard, N., Tucker, G. S., Weiland, J. L., Wollack, E., and Wright, E. L.: 2007, Astrophys. J., Suppl. Ser. 170, 377 Springel, V., White, S. D. M., Jenkins, A., Frenk, C. S., Yoshida, N., Gao, L., Navarro, J., Thacker, R., Croton, D., Helly, J., Peacock, J. A., Cole, S., Thomas, P., Couchman, H., Evrard, A., Colberg, J., and Pearce, F.: 2005, Nature 435, 629 Stanford, S. A., Eisenhardt, P. R., Brodwin, M., Gonzalez, A. H., Stern, D., Jannuzi, B. T., Dey, A., Brown, M. J. I., McKenzie, E., and Elston, R.: 2005, Astrophys. J., Lett. 634, L129 Stanford, S. A., Elston, R., Eisenhardt, P. R., Spinrad, H., Stern, D., and Dey, A.: 1997, Astron. J. 114, 2232 Stanford, S. A., Holden, B., Rosati, P., Eisenhardt, P. R., Stern, D., Squires, G., and Spinrad, H.: 2002, Astron. J. 123, 619 Stanford, S. A., Holden, B., Rosati, P., Tozzi, P., Borgani, S., Eisenhardt, P. R., and Spinrad, H.: 2001, Astrophys. J. 552, 504 Stanford, S. A., Romer, A. K., Sabirli, K., Davidson, M., Hilton, M., Viana, P. T. P., Collins, C. A., Kay, S. T., Liddle, A. R., Mann, R. G., Miller, C. J., Nichol, R. C., West, M. J., Conselice, C. J., Spinrad, H., Stern, D., and Bundy, K.: 2006, Astrophys. J., Lett. 646, L13 St¨ocker, H.: 1998, Taschenbuch der Physik, Frankfurt am Main: Verlag Harri Deutsch Stott, J. P., Smail, I., Edge, A. C., Ebeling, H., Smith, G. P., Kneib, J.-P., and Pimbblet, K. A.: 2007, Astrophys. J. 661, 95 Strazzullo, V., Rosati, P., Stanford, S. A., Lidman, C., Nonino, M., Demarco, R., Eisen- hardt, P. E., Ettori, S., Mainieri, V., and Toft, S.: 2006, Astronomy & Astrophysics 450, 909 Struble, M. F. and Rood, H. J.: 1999, Astrophys. J., Suppl. Ser. 125, 35 Sunyaev, R. A. and Zeldovich, Y. B.: 1972, Comments on Astrophysics and Space Physics 4, 173 Tanaka, M., Kodama, T., Kajisawa, M., Bower, R., Demarco, R., Finoguenov, A., Lidman, C., and Rosati, P.: 2007, ArXiv Astrophysics e-prints, astro-ph/0703135 Thomas, D., Maraston, C., Bender, R., and Mendes de Oliveira, C.: 2005, Astrophys. J. 621, 673 Tinsley, B. M.: 1978, Astrophys. J. 222, 14 Tonry, J. and Davis, M.: 1979, Astron. J. 84, 1511 Valtchanov, I., Pierre, M., and Gastaud, R.: 2001, Astronomy & Astrophysics 370, 689 Valtchanov, I., Pierre, M., Willis, J., Dos Santos, S., Jones, L., Andreon, S., Adami, C., Altieri, B., Bolzonella, M., Bremer, M., Duc, P.-A., Gosset, E., Jean, C., and Surdej, J.: 2004, Astronomy & Astrophysics 423, 75 van Dokkum, P. G. and Franx, M.: 1996, MNRAS 281, 985 van Dokkum, P. G., Franx, M., Kelson, D. D., Illingworth, G. D., Fisher, D., and Fabricant, D.: 1998, Astrophys. J. 500, 714 Vikhlinin, A., Burenin, R., Forman, W. R., Jones, C., Hornstrup, A., Murray, S., and Quintana, H.: 2007, in H. B¨ohringer, G. W. Pratt, A. Finoguenov, and P. Schuecker (eds.), Heating versus Cooling in Galaxies and Clusters of Galaxies, p. 48 264 BIBLIOGRAPHY

Vikhlinin, A., Forman, W., Jones, C., and Murray, S.: 1995, Astrophys. J. 451, 542 Vikhlinin, A., McNamara, B. R., Forman, W., Jones, C., Quintana, H., and Hornstrup, A.: 1998, Astrophys. J. 502, 558 Voit, G. M.: 2005, Reviews of Modern Physics 77, 207 von der Linden, A., Best, P. N., Kauﬀmann, G., and White, S. D. M.: 2007, MNRAS 379, 867 Wetterich, C.: 1988, Nuclear Physics B 302, 668 White, S. D. M.: 1976, MNRAS 174, 19 White, S. D. M.: 2007, ArXiv Astrophysics e-prints, astro-ph/07042291 White, S. D. M. and Rees, M. J.: 1978, MNRAS 183, 341 Worthey, G.: 1994, Astrophys. J., Suppl. Ser. 95, 107 Zhang, Y.-Y.: 2005, Ph.D. thesis, Ludwig-Maximilian University Munich Zibetti, S., White, S. D. M., Schneider, D. P., and Brinkmann, J.: 2005, MNRAS 358, 949 Zwicky, F.: 1937, Astrophys. J. 86, 217 List of Figures

2.1 BulletCluster...... 5 2.2 ThermalX-raySpectra...... 6 2.3 ClusterMassProﬁles ...... 12 2.4 ClusterTemperatureProﬁles...... 13 2.5 Color-Magnitude Diagram for the Lynx Supercluster ...... 22 2.6 Simulated Cluster Ellipticals ...... 24 2.7 MassAssemblyHistoryofBCGs...... 27 2.8 BCGModelEvolution ...... 28 2.9 BCGStellarMassDistribution ...... 28

3.1 CosmologicalDistances...... 36 3.2 Fluxes and Comoving Volumes ...... 36 3.3 CosmicTime ...... 37 3.4 ClusterFormationSimulations...... 42 3.5 HaloNumberDensity...... 46 3.6 LSSSimulations...... 47

4.1 X-ray Cluster Survey Comparison ...... 59 4.2 ImportantLocalSurveyResults ...... 60 4.3 X-ray Cluster Luminosity Function ...... 60 4.4 NORAS2GalaxyCluster ...... 61 4.5 NORAS2GalaxyClusterSpectra ...... 61 4.6 ROSATClusterNumberCounts...... 64 4.7 DistantClusterRDCS1252.9-2927 ...... 66 4.8 ConeDiagramofKnownGalaxyClusters...... 66

5.1 XDCPSurveyStrategy...... 69 5.2 XMMUJ2235Properties...... 74

6.1 XMM-Newton...... 78 6.2 XMM Vignetting and PSF Functions ...... 82 6.3 XMMPSFShape...... 82 6.4 XMM-NewtonArchivalFields ...... 84 6.5 XMM-NewtonSkyCoverage...... 84

265 266 LIST OF FIGURES

6.6 OptimizedDetectionBands ...... 88 6.7 DetectionBandSchemes ...... 88 6.8 X-rayPipelineFlowChart ...... 89 6.9 X-rayFlareCleaning ...... 92 6.10 X-raySourceDetectionProducts ...... 95 6.11 WaveletSourceDetection ...... 99 6.12 X-ray-OpticalOverlays ...... 102 6.13 SourceScreeningFlowChart...... 105 6.14FlaggingScheme ...... 109 6.15 ClassificationPrototypes ...... 109 6.16 DSSIdentificationLimit ...... 111 6.17 Redshift Distribution of Identified COSMOS Clusters ...... 113 6.18 SpaceWeatherSeasons...... 115 6.19 XDCPSurveyFieldStatistics ...... 116 6.20 Extent Likelihood versus X-ray Source Counts ...... 118 6.21 ComparisonofDetectionSchemes ...... 119 6.22 Cluster Detections versus Off-axis Angle ...... 119

7.1 CMDModels ...... 122 7.2 Redshift Uncertainties of Red Sequence Methods ...... 129 7.3 OMEGA2000InstrumentResponse ...... 130 7.4 NIR Science Reduction Pipeline Flow Chart ...... 136 7.5 NIR Science Reduction Steps ...... 137 7.6 FaintObjectOptimization ...... 144 7.7 Advanced Reduction Schemes ...... 145 7.8 Abell383Field ...... 149 7.9 NIRNumberCounts ...... 151 7.10 NIRPhotometricAnalysis ...... 152 7.11 PhotometricCalibration ...... 159

8.1 SpectralSkyFeatures...... 167 8.2 SpectroscopyReductionFlowChart...... 169

9.1 XDCPSkyCoverage ...... 174 9.2 Raw log N–log S ofXDCPClusters ...... 176 9.3 Extent Properties of XDCP Candidates ...... 177 9.4 Calar Alto Imaging Cluster Sample ...... 182

10.1 Z–HEvolutionoftheRed-Sequence ...... 188 10.2 Formation Redshift of Ellipticals ...... 189 10.3 CalibrationClustersA ...... 192 10.4 CalibrationClustersB ...... 193 10.5 Spectroscopic Conﬁrmation of XMMU J0104.4-0630 ...... 194 10.6 CMDofXMMUJ0104.4-0630 ...... 196 LIST OF FIGURES 267

10.7 Galaxy Overdensities in the Field of XMMU J0104.4-0630 ...... 197 10.8 Galaxy Populations of XMMUJ0104.4-0630 ...... 198

11.1BCGSelection...... 202 11.2BCGHubblediagram...... 204 11.3 BCGAbsoluteMagnitudes...... 205 11.4 BCG K-correction ...... 205 11.5 BCGLuminosityEvolution ...... 208 11.6 BCGAssemblyModels...... 209 11.7 High-redshiftBCGgallery...... 216 11.8 Low-redshiftBCGgallery...... 217 11.9 High-Redshift Cluster Gallery A...... 221 11.10High-Redshift Cluster Gallery B...... 222

12.1 South Pole Telescope and eROSITA ...... 224 12.2 VADERSpacecraftandSurveyRegion ...... 229 268 LIST OF FIGURES List of Tables

2.1 ClusterScalingRelation ...... 15

3.1 Cosmological Concordance Parameters ...... 37 3.2 Cosmological Tests with Galaxy Clusters ...... 55

5.1 PropertiesofXMMUJ2235.3-2557 ...... 73 5.2 ThesisDataOverview ...... 75

6.1 XMM-NewtoninaNutshell ...... 81

7.1 XDCPFilters ...... 123 7.2 ApparentMagnitudeEvolution ...... 124 7.3 SummaryofImagingMethods ...... 129 7.4 OMEGA2000Characteristics...... 131 7.5 NIRDataOverview...... 134

8.1 SpectralGalaxyFeatures...... 165

9.1 Spectroscopically Conﬁrmed z>1X-rayClusters ...... 185

10.1 X-rayProperties ofXMMUJ0104.4-0630 ...... 196

12.1 ExpectedVADERResults ...... 226

A.1 XDCPSurveyFields ...... 231

269 270 LIST OF TABLES Acknowledgements

It is my pleasure to thank Hans B¨ohringer for being a superb supervisor, for giving me valuable advice when needed, for encouraging new projects and initiatives, for support- ing new ideas and approaches, for sponsoring frequent travel, and for always spreading a positive, open attitude.

This thesis is dedicated to the late Peter Schuecker, my mentor and advisor, whose enthu- siasm and love for the beauty of the Universe has been truly inspiring and will continue to live on.

I am indebted to many other people who contributed to this thesis either scientifically, financially, with moral support, or just by making MPE a great place to work. I would like to thank Prof. Gregor Morfill for his support over the years, his bottles of wine for late work sessions, and for being a humorous office neighbor. I thank Wolfgang Voges for being a great thesis committee member and for his encouragement.

Furthermore, I am grateful to the whole XDCP team. Special thanks go to Chris Mullis for getting me started, Georg Lamer for his X-ray help, and Joana Santos for helping with the observations. I am also indebted to the Calar Alto staﬀ for their support, with additional thanks to Aurora Simionescu and Filiberto Braglia for their participation. I have also enjoyed the CTIO observing run and the ongoing work with Joe Mohr. Special thanks go to Hermann-Josef R¨oser for the support of OMEGA2000, the active exchange of ideas and codes, and his advice. Thanks also to Harald Baumgartner and Joachim Paul for their essential technical support over the years.

My office mate Gabriel Pratt deserves special thanks for putting up with me, for the convivial work atmosphere, his X-ray support, and all the proofreading. I would also like to thank Alexis Finoguenov for sharing his expertise, data, and money with me. Very warm final thanks go to Daniele Pierini for providing all the models, galaxy expertise, guidance, and proofs, and to Martin Mühlegger for his careful reading and cheerful personality.